The mobility sector plays a foundational role in the economy of the European Union (EU), but it faces growing pressure to adapt to a rapidly changing landscape. Artificial intelligence (AI) is emerging as a key enabler of smarter, more sustainable and more resilient mobility systems to address these challenges. This chapter maps areas in which AI is already being deployed or holds strong potential in AI mobility systems, and analyses three of these application fields in depth: automated driving, public transport, and fleet management (freight transport). The chapter is based on a literature review and interviews and workshops with EU business associations and enterprises between December 2024 and April 2025.
Progress in Implementing the European Union Coordinated Plan on Artificial Intelligence (Volume 2)
Uptake in High-Impact Sectors
Abstract
The mobility sector plays a foundational role in the economy of the European Union (EU), enabling the movement of people and goods; supporting industrial supply chains; and connecting cities, regions and markets. However, the sector is under growing pressure to adapt to a rapidly changing landscape. Rising congestion, ageing infrastructure, the need to decarbonise in line with climate targets, growing global competition and supply chain dependencies, along with increasing demands for safety and efficiency, all highlight the urgent need for transformation.
Digitalisation offers a pathway to address many of these challenges. Artificial intelligence (AI) is emerging as a key enabler of smarter, more sustainable and more resilient mobility systems. AI can help optimise traffic flows in real time, support predictive maintenance (PdM) and planning of infrastructure, reduce emissions through intelligent route planning and enable more efficient use of renewable electricity and grid resources.
This chapter provides an overview of how the EU mobility sector is using AI and identifies what is needed to scale these technologies more widely and responsibly in the region. It begins with a broad mapping of application areas in which AI is already being deployed or holds strong potential – from vehicle automation to freight co‑ordination and emissions tracking.
The chapter is structured in two parts:
Overview of the EU mobility sector and the strategic role of AI: an outline of the structural and economic make-up of the sector presents quantitative data on AI adoption. It highlights the types of AI tools being used by enterprises; differences in uptake across Member States; the availability of AI-related skills; and trends in venture capital (VC) investment in AI-based mobility solutions. The chapter also includes a broad overview of AI applications across the mobility landscape, offering a functional classification of use-cases based on their operational purposes, technical characteristics and barriers to deployment reported in literature.
Spotlight on selected AI application fields in mobility: the second part provides an in-depth analysis of three application fields in the mobility sector: i) automated driving; ii) public transport systems; and iii) fleet management and heavy transport logistics. Each section begins with a synthesis of insights from academic and policy literature, followed by findings drawn from stakeholder interviews and expert workshops. For each application field, the analysis identifies key barriers and challenges, as well as targeted policy recommendations to support responsible, effective and scalable adoption of AI technologies.
Methodological considerations are discussed in Chapter 1. A summary of the key recommendations is provided next.
Data availability and access
Foster secure, interoperable data sharing to promote EU-level data pooling and address issues of data ownership, privacy and competition.
Invest in data standardisation and integration to support the seamless exchange across public transport, automated vehicles and multimodal logistics,
Build collaborative data platforms to provide access to clean and standardised data and facilitating co‑operation between public and private actors.
Infrastructure and connectivity
Upgrade infrastructure to retrofit roads, terminals and facilities to support safe, efficient AI-enabled use-cases.
Expand digital and physical AI-ready digital and physical infrastructure to enable connected and automated mobility applications.
Bridge the urban-rural divide to equip smaller cities, ports and terminals with the infrastructure needed for AI adoption.
AI solutions, software and interoperability
Promote open-source, modular and interoperable AI ecosystems to reduce vendor lock-in, support adoption by small and medium-sized enterprises, and stimulate innovation across the transport sector.
Enable experimentation through pilots and virtual testing to allow public and private actors to test AI solutions in controlled environments, evaluate their impact and reduce investment risk before large-scale deployment.
Foster development of sector-specific large language models (LLMs) tailored to the mobility and transport sector, leveraging proprietary European datasets to build competitive and trustworthy AI solutions.
Regulatory frameworks and testing environments
Ensure the AI Act and related regulations are accompanied by clear, practical guidance for transport sector stakeholders.
Establish national regulatory sandboxes and cross-border testbeds where cities, regions and logistics operators can experiment with AI solutions, waiving compliance burdens in controlled settings, and facilitating innovation while safeguarding public interest.
Skills, talent and collaboration
Invest in AI workforce development and upskilling to address skills gaps across both public and private actors in transport and mobility sectors.
Strengthen public-private-academic innovation partnerships to co-develop AI applications addressing operational challenges in the mobility sector.
Promote peer learning and public trust in AI applications, address societal concerns (around privacy, safety and employment), and ensure inclusive adoption of AI-driven mobility solutions across Member States.
This section outlines the structural and economic make-up of the sector, presenting quantitative data on adoption of AI. It highlights the types of AI tools being used by enterprises; differences in uptake across Member States; the availability of AI-related skills; and trends in VC investment in AI-based mobility solutions. It also includes a broad overview of AI applications across the mobility landscape, offering a functional classification of use-cases based on their operational purposes, technical characteristics and factors affecting deployment.
Key characteristics of the mobility sector
Sectoral make-up
The mobility sector is a pillar of the economy of the European Union (EU), playing a fundamental role in trade, employment and connectivity across Member States. It serves as the backbone of European economic activity by enabling access to people, opportunities and goods; supporting industrial supply chains; and facilitating international trade and transport.
Through its various modes, the transport sector is a major contributor to the EU economy. In 2023, the EU transport sector accounted for approximately 5% to the region’s gross domestic product (GDP). It also employed about 6.2 million people, underscoring its essential role in economic activities and supply chains (Eurostat, 2025[1]). The sector comprises both freight and passenger transport, operating in a wide range of services including land (road, rail, urban roads), waterborne (inland waterways, coastal, ocean) and air transport.
Among freight transport modes, maritime transport dominates in terms of weight transported (measured in tonne-kilometres). Maritime transport accounted for 67.8% of total freight transport activity in 2022, followed by road transport (24.9%), rail (5.5%), inland waterways (1.6%) and air transport (0.2%) (Figure 5.1). Sea transport plays an especially significant role in international trade, handling 47% of the total value of goods exchanged between EU and non-EU countries. In that same year, air transport contributed 26.2% to exports and 17.4% to imports, while road transport accounted for 24.1% of exports and 18.7% of imports (Eurostat, 2024[2]).
The sector’s extensive infrastructure comprises over 5 million kilometres (km) of paved roads, approximately 73 200 km of motorways and a rail network spanning more than 200 000 km (Eurostat, 2025[3]). Investment in transport infrastructure remains a key priority for the European Union, with a strong emphasis on modernisation, sustainability and digitalisation. In 2022, general government expenditure on transport infrastructure stood at 2.2% of GDP, slightly above the 2.1% recorded in 2012. Expenditure fluctuated between 2.0% and 2.1% from 2012 to 2019 before rising to 2.3% in 2020 and 2021, reflecting increased public support for transport operators during the COVID-19 pandemic. By 2022, the ratio had slightly declined to 2.2%, reflecting economic stabilisation and a return to pre-pandemic spending trends (Eurostat, 2025[1]).
Investment in transport equipment has also followed a dynamic trajectory. Gross fixed capital formation in transport equipment relative to GDP rose from 1.5% in 2013 to a peak of 2.0% between 2017 and 2019. However, the COVID-19 crisis led to a decline, with the ratio dropping to 1.7% in 2020, 2021 and 2022. It rebounded to 1.9% in 2023, indicating renewed investment in fleet modernisation and infrastructure upgrades (Eurostat, 2025[1]).
The EU mobility sector is undergoing a profound transformation, driven in part by environmental imperatives. One of the most pressing challenges is the need to decarbonise transport, which accounts for nearly one-quarter of EU greenhouse gas (GHG) emissions (EEA, 2025[4]). Achieving the targets set out in the European Green Deal (European Commission, 2019[5]) and the Sustainable and Smart Mobility Strategy (European Commission, 2020[6]) requires large-scale adoption of clean technologies, such as electric vehicles (EVs), alternative fuels and energy-efficient infrastructures. It also requires significant modal shifts towards rail and public transport.
Percentage, based on tonne-kilometres
Note: Countries report maritime transport data based on passengers or freight carried between pairs of ports. Eurostat calculates tkm/pkm for each port pair using a distance matrix and then “territorialises” these values by allocating them proportionally to countries according to the distance travelled within their exclusive economic zones (EEZ). It does not consider distances beyond the EEZ. While some aspects of the methodology are not fully detailed, maritime transport accounts for over 80% of global goods trade. This suggests the data may include flows with an EU origin or destination, not only intra-EU transport.
Source: Eurostat (2024[2]), “Freight transport statistics – modal split”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Freight_transport_statistics_-_modal_split.
In addition to decarbonisation, the sector faces a series of persistent and emerging challenges related to technological disruption and shifting economic and social dynamics. These include ageing infrastructure, urban congestion and fragmentation across Member States. This is especially the case in areas such as rail interoperability, cross-border ticketing and EV charging infrastructure. The green and digital transitions are also revealing skill shortages in the transport workforce, underscoring the urgent need for greater investment in digital infrastructure, cybersecurity and innovation ecosystems.
Furthermore, global competition is intensifying in strategic areas such as battery supply chains and AI-enabled mobility services. This is raising concerns about technological sovereignty and long-term competitiveness in key transport domains, such as the automotive sector (European Commission, 2025[7]).
In this context, digitalisation and AI are emerging as critical enablers of transformation across the EU mobility sector. These technologies could reshape how transport systems are planned, managed and experienced, offering solutions to some of the sector’s most pressing challenges. By increasing the efficiency, sustainability and resilience of transport systems and mobility services, digital technologies and AI applications can advance decarbonisation, improve infrastructure planning and maintenance, enhance user experience and optimise resource use.
AI use-cases
AI can play a key role in making transport systems smarter and more efficient across a wide range of uses. It helps optimise routes and schedules by adapting to traffic conditions in real time, improving the flow of everything from buses to freight. AI also forecasts travel demand by analysing patterns and events, helping transport planners stay ahead. In maintenance, it spots early signs of wear or failure in vehicles and infrastructure, reducing breakdowns and delays. AI-powered cameras and sensors monitor road and rail conditions, while traffic flow is improved through smart signal control and congestion prediction. It also supports automated vehicles (AVs) with tasks like navigation, obstacle detection and motion control. In logistics, AI helps co‑ordinate complex, multimodal freight systems and manage autonomous fleets. It contributes to better energy use by planning EV charging and optimising fuel efficiency. In addition, AI enhances safety by detecting risks early and supports a better user experience with personalised travel advice and accessibility features.
To provide a structured overview of the diverse AI applications emerging in the mobility sector, this section categorises use-cases according to their primary functions, aligning them with the OECD Framework for the Classification of AI systems (OECD, 2022[8]). This mapping is intended to illustrate the breadth of AI’s potential across modes and transport functions, from PdM and traffic signal control to automated fleet co‑ordination and cybersecurity. Table 5.1 summarises these applications, describing their functionality, learning methodologies and associated challenges.
While AI has a wide range of applications across mobility sectors, three key areas stand out for their transformative potential: i) automated driving; ii) public transport systems; and iii) fleet management (freight transport). These areas encompass the most significant advancements in AI-driven mobility and are critical for the evolution of smart transport networks. This chapter delves into these three use-cases, exploring their technological foundations, real-world applications and the challenges that must be addressed to fully leverage the capabilities of AI in mobility in the European Union.
In addition to these domain-specific AI applications, there is a growing interest in the role of foundation models, including both small and large language models (LLMs), as well as other generative AI tools, in the mobility sector. These models could enable cross-cutting functionalities that support multiple use-cases by leveraging natural language processing, data summarisation and multimodal integration capabilities. Throughout the three key use-cases explored in this chapter – automated driving, public transport and AI for fleet management (freight transport) – such models are being investigated as complementary tools to enhance perception, prediction, co‑ordination and user interaction. Their potential contributions range from powering conversational interfaces and multimodal information systems in public transport to facilitating automated document processing and decision support in logistics operations.
Level of AI uptake in the transport sector in the European Union
The level of AI uptake in the EU transport sector is increasing but remains uneven. As investment in modernising fleets and transport infrastructure accelerates, attention is also shifting towards the digital transformation of the sector, especially adoption of emerging technologies such as AI. Yet, despite its strategic importance, AI uptake in the transport sector is characterised by highly uneven diffusion (Figure 5.2).
As a percentage of enterprises with ten or more employees
Notes: “All activities” include all economic sectors except agriculture, forestry and fishing; mining and quarrying; and the financial sector. The percentage of firms using AI in a given industry and year is defined as the number of firms using at least one AI technology relative to all firms in the specific industry. Includes only firms with at least ten employees. AI technologies include performing analysis of written language (text mining), generating written or spoken language (natural language generation), automating different workflows or assisting in decision making (AI-based software robotic process automation), converting spoken language into machine-readable format (speech recognition), identifying objects or persons based on images (image recognition, image processing), machine learning (e.g. deep learning) for data analysis and, enabling physical movement of machines via autonomous decisions based on observation of surroundings (autonomous robots, self-driving vehicles, autonomous drones). See Eurostat (2025[9]) for further details on the data obtained through the “EU survey on ICT usage and e‑commerce in enterprises”.
Source: Eurostat (2025[3]), “Statistics explained: Use of artificial intelligence in enterprises”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises.
The information and communication sector leads AI adoption in the European Union, with over 40% of enterprises reporting the use of AI technologies in 2024. This is followed by the professional, scientific and technical activities sector (just above 30%) and real estate activities (around 20%). These sectors are typically more digitised and have business models well aligned with AI integration, often involving data-driven services, automation and predictive analytics.
In contrast, the transport and storage sector ranks among the lowest adopters of AI technologies, with only 8.1% of enterprises reporting AI use in 2024. This is not only below typically more digitalised sectors but also lags traditionally less tech-intensive sectors like manufacturing (10.6%). AI uptake in transport is more comparable to sectors such as water supply and environmental management, where adoption also remains modest.
Several structural and sector-specific factors help explain this comparatively low adoption rate. First, transport systems rely on capital-intensive, long-life assets such as vehicles, signalling infrastructure and control systems. These are costly to retrofit or replace solely to incorporate the latest digital technologies. Second, many promising AI applications – such as traffic optimisation, PdM or intermodal co‑ordination – depend on complex network effects. These require synchronised action across multiple independent actors, including infrastructure providers, operators and public authorities. Third, transport is a safety-critical sector where societal concerns and legal requirements constrain the use of “black box” AI systems, particularly in high-risk applications like automated driving or maintenance automation.
These structural features are especially pronounced in public transport and infrastructure services, where public authorities play a central role in operations, oversight and service delivery. In urban and ground passenger transport, for instance, public authorities typically operate services. Otherwise, these services are outsourced under strict obligations that prioritise public safety, accessibility and equity. These obligations may constrain the use of AI tools by private contractors. Firm-level AI adoption statistics may not fully capture the extent or nature of AI use in the transport domain, especially where deployment is mediated through public procurement, delegated services or infrastructure authorities.
Compared to other sectors, transport has lower adoption of AI across most use-cases. Among companies that already use AI, enterprises in the transport and storage sector most frequently apply it for business administration processes or management (27.51%). This is followed closely by information and communication technology (ICT) security (22.38%) and accounting, controlling or finance management (22.65%) (Figure 5.3). Compared to the economy-wide average, the transport sector demonstrates lower adoption of AI across most use-cases. The most striking gap is in marketing and sales, where 34.08% of all enterprises use AI, compared to only 24.54% in transport. Similarly, AI for research and development or innovation activities is used by 18.59% of all enterprises, but by just 9.74% in transport. The only area where transport outpaces the overall average is logistics, with 18.78% of transport enterprises using AI compared to 6.12% across all sectors, reflecting the strategic importance of logistics optimisation in this sector.
Within the transport sector, some areas adopt AI more than others. The most used AI function in this sector is text mining, with just under 4% of enterprises adopting it compared to nearly 8% across all industries (Figure 5.4). Natural language generation, speech recognition, and marketing or sales, are each implemented by around 2‑3% of transport enterprises. However, adoption in critical operational areas such as production processes and autonomous systems (such as self-driving vehicles and drones) remains low, with each used by less than 1% of enterprises. Not only does this situation highlight a significant gap in AI adoption between the transport sector and the broader economy, but it also reflects the uniquely high-risk nature of many transport applications. Vehicle operation in open traffic, for example, is subject to strict safety, liability and certification requirements. Regulatory frameworks have yet to fully determine whether these high-risk AI systems can be deployed safely at scale. As a result, there is potential for growth in operational AI applications that could enhance efficiency and automation. However, such progress will depend on the evolution of safety standards, oversight mechanisms and public trust in AI-supported mobility systems.
As a percentage of enterprises with ten or more employees
Note: “All activities” include all economic sectors except agriculture, forestry and fishing; mining and quarrying; and the financial sector.
Source: Eurostat (2025[3]), “Statistics explained: Use of artificial intelligence in enterprises”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises.
As a percentage of enterprises with ten or more employees
Note: “All activities” include all economic sectors except agriculture, forestry and fishing; mining and quarrying; and the financial sector.
Source: Eurostat (2025[3]), “Statistics explained: Use of artificial intelligence in enterprises”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises.
EU-wide trends mask significant variation between Member States
Some countries emerge as leaders in AI adoption while others lag (Figure 5.5).
As a percentage of enterprises with ten or more employees
Source: Eurostat (2025[3]), “Statistics explained: Use of artificial intelligence in enterprises”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises.
In 2024, Belgium (23%), Denmark (21.1%) and Malta (20.6%) emerged as clear frontrunners, each with over 20% of enterprises using AI technologies. This disparity may partly reflect sectoral specialisation. The elevated adoption rates in Belgium and Malta, for example, are likely due in part to the prominence of port and logistics operations. Meanwhile, the strong maritime and shipping industry in Denmark, including AI adoption by major players such as Maersk, likely contribute to its elevated adoption rate.
Close behind, Slovenia (19.0%), Sweden (15.5%), Luxembourg (14.5%) and Austria (13.5%) also demonstrated strong adoption, well above the EU27 average of 8.1%. Several other countries, including Greece (11.6%), Germany (11.5%) and the Netherlands (11.0%), slightly exceeded the EU average, indicating solid – though less striking – progress in integrating AI into their transport sectors.
By contrast, a significant number of Member States remained below the EU average. Countries like Croatia (7.8%), Ireland (7.9%), Cyprus (7.3%) and Finland (7.0%) recorded moderate adoption, while France (5.3%), Italy (5.2%) and Portugal (4.9%) lagged further behind despite their sizeable economies. At the lower end of the spectrum, Romania (2.7%), Poland (2.5%) and Hungary (4.0%) reported the weakest uptake.
Notably, adoption rates have generally risen across the board since 2021. Between 2021 and 2024, adoption rates have climbed from 5.2% to 8.13% overall for the EU27. Countries like Sweden have made significant leaps – from 3.7% to 15.5% during the same period (Eurostat, 2025[3]). Similarly, AI talent concentration as reported by LinkedIn members in transport, logistics, supply chain and storage, increased between 2018 and 2024. However, the share of workers with AI skills remains low in the sector (Figure 5.6). This trend underscores increasing awareness of how AI can enhance logistics, automation and operational efficiency, although the pace of integration remains uneven.
Notes: This chart shows the concentration of LinkedIn members with at least two AI engineering skills or who perform an AI occupation per country, industry and in time. Please see the methodological note for more information.
Source: (OECD.AI, 2025[10]), AI talent concentration by country and industry, calculations based on data from LinkedIn Economic Graph, last updated 2025-04-07, https://oecd.ai/.
Global adoption trends of AVs are expected to follow divergent trajectories across regions, according to the World Economic Forum (2025[11]). These projections differ significantly by automation level, which ranges from basic driver assistance (L1) to full autonomy (L5). In brief, L1 involves assisted driving features such as lane-keeping or adaptive cruise control. L2 offers partial automation with combined steering and speed control. L3 enables conditional automation where the system takes over under defined conditions (e.g. traffic jam pilots). L4 allows fully autonomous driving within specific operational design domains (ODDs) and L5 refers to full autonomy in all driving conditions.
The People’s Republic of China (hereafter “China”) is projected to lead the uptake of personal vehicle automation, followed by the United States, with Europe and Japan progressing at a slower but steady pace. By 2035, the share of new car sales equipped with advanced driver-assistance systems (ADAS) (L2+) is expected to be significantly higher in China. This projected increase is driven by strong consumer acceptance and rapid advances by original equipment manufacturers (OEMs) in China. Levels 3 and 4 (L3/L4) vehicles are also expected to appear in China, the United States and Europe by this time. However, other regions may face economic, regulatory and technological constraints that delay higher levels of automation. India, for instance, is forecasted to retain a predominance of non-automated vehicles (L0) in 2035 due to lower purchasing power and challenging road environments. The country’s adoption of L2 technologies is likely to bypass intermediate L1 solutions as the technology matures.
In this evolving global landscape, European start-ups are playing a dynamic role by developing AI-driven mobility solutions and advancing AV technologies tailored to local needs and market conditions. Several companies are pioneering AV technologies. In the Netherlands, 2getthere is developing automated shuttle systems for public transport, while Einride in Sweden is known for its electric, self-driving trucks for freight mobility. In Finland, Sensible 4 specialises in automated driving software tailored for harsh weather conditions, while Auve Tech in Estonia offers electric automated transport systems. Other start-ups focus on intelligent communication and infrastructure. Veniam in Portugal and Commsignia in Hungary develop AI solutions for V2V and V2I communication. FlowX in Romania and Transmetrics in Bulgaria provide platforms for process optimisation and logistics forecasting. More established firms such as Bolt (Estonia) and VivaDrive (Poland) apply AI to optimise ride-hailing services and fleet electrification, respectively. In addition to investments in AI start-ups, infrastructure managers and public authorities are investing in embedding AI into infrastructure management, service delivery and trip planning.
However, VC investment in AI start-ups focused on mobility and AVs has fluctuated significantly over the past decade. VC has had moments of strong growth, notably in 2022 where total VC funding reached USD 277 million. However, these start-ups have consistently received only a small portion (between 1‑3%) of total VC investment in AI within the European Union. This highlights the sector’s relative underinvestment compared to other AI applications (Figure 5.7). This share is comparable to that of the United Kingdom. However, it is significantly lower than in the United States, where mobility-related AI start‑ups have attracted on average around 12% of total AI VC funding in recent years (OECD.AI, 2025[12]).
Note: Please see the methodological note for more information.
Source: OECD.AI (2025[12]), Total VC investments in AI by country and industry, calculations based on data from Preqin, last updated 2025-02-18, https://oecd.ai/.
Select key EU legislative acts applying to AI in the mobility sector
As the mobility sector undergoes rapid digital transformation, the European Union has introduced a set of interlinked legislative acts to ensure the safe, secure and trustworthy deployment of AI across transport systems. These legislative acts govern how AI systems are designed, validated, used and monitored, balancing innovation support with risk mitigation.
EU AI Act: The AI Act provides the overarching horizontal framework for the regulation of AI in the European Union. It classifies AI systems into four risk categories and imposes obligations proportionate to their risk level.
Intelligence Transport System Directive: The ITS Directive (Directive (EU) 2023/2661 amending 2010/40/EU) (European Parliament & Council, 2023[13]) governs deployment of intelligent transport systems (ITS) across the European Union. It was revised in 2023 to reflect the increased use of AI and data-driven mobility services.
General Safety Regulation: The General Safety Regulation (Regulation (EU) 2019/2144) introduces mandatory safety technologies for vehicles placed on the EU market. From July 2024, it has required deployment of several advanced safety features, many of which are AI-enabled.
Vehicle-Type-Approval Regulation: The Vehicle Type-Approval Regulation (Regulation (EU) 2018/858) regulates how vehicles – including those with autonomous functionalities – are tested and approved before entering the EU market.
Electronic Freight Transport Information Regulation (eFTI): The Electronic Freight Transport Information (Regulation (EU) 2020/1056) facilitates the digital exchange of freight transport information between businesses and authorities.
General Data Protection Regulation (GDPR): While not mobility-specific, the GDPR has critical implications for all AI systems handling personal data.
Delegated Regulation on Multimodal Travel Information Services: The Delegated Regulation on Multimodal Travel Information Services (EU) 2017/1926 mandates the availability and accessibility of data to support multimodal travel information services (MMTIS) across the European Union.
The following sections explore in depth three selected use-cases that exemplify the transformative potential of AI in mobility.
Automated driving
Automated driving is among the most technically complex and ambitious applications of AI in mobility, integrating AI-driven perception, localisation, mapping, decision making and motion control. AI enables vehicles to interpret their surroundings, anticipate road conditions and make adaptive driving decisions. ADAS are already widely deployed in commercial vehicles, supporting tasks such as motorway driving, parking and collision avoidance. However, fully autonomous systems remain largely confined to pilot projects and controlled environments (Ma et al., 2020[14]; Noviati et al., 2024[15]). However, ongoing advancements in AI, sensor fusion and vehicle-to-everything (V2X) communication are expected to further improve automation capabilities, particularly in freight transport and urban mobility.
Main use-cases reported in literature
Academic literature highlights key AI applications in perception, decision making and motion control. AI‑powered perception systems process sensor data from LiDAR, radar and cameras, enabling vehicles to detect objects and interpret road conditions (Khayyam et al., 2019[16]; Muhammad et al., 2021[17]). Reinforcement learning models and hybrid AI approaches allow AVs to adjust speed, lane positioning and route selection based on real-time conditions (Atakishiyev et al., 2024[18]; Garikapati and Shetiya, 2024[19]). Motion control systems, incorporating model predictive control and deep neural controllers, enhance vehicle manoeuvrability and response to dynamic environments (Parekh et al., 2022[20]; Reda et al., 2024[21]). While these advances improve reliability, their safe integration into public road networks remains an area of uncertainty, active research and regulatory discussion. Large-scale demonstrations and cross‑border test beds are validating several of these advances. This process is supported by the EU Connected, Cooperative and Automated Mobility (CCAM) Partnership, which aims to accelerate the transition from research prototypes to real-world applications (European Commission, 2024[22]).
Perception and localisation
AVs rely on AI-powered perception systems to process data from a combination of sensors, including LiDAR, radar, cameras and ultrasonic detectors. These sensors work together to detect and classify objects, identify road users and track lane markings in real time (Khayyam et al., 2019[16]). Deep- learning models, particularly convolutional neural networks, have significantly enhanced the ability of AVs to recognise and predict the movement of objects with high accuracy.
Interpretation of data goes beyond sensing and perceiving. Sensing involves acquiring data from the environment based on sensors’ capabilities. Perception implies the isolation of what in the field of data matches a certain threshold of importance. Finally, interpreting refers to assigning a meaning to the perceived phenomena to extract meaningful information.
To ensure a robust interpretation of the environment, AI-powered sensor fusion techniques, such as Kalman filters and Bayesian networks, integrate data from multiple sources, improving situational awareness even in challenging weather conditions (Muhammad et al., 2021[17]). Recent advances in self-supervised learning are further refining these perception models. These allow AI to learn from vast amounts of unlabelled data, reducing dependence on manual annotations while improving detection accuracy (Cui et al., 2024[23]).
Beyond static object detection, AI-enhanced Simultaneous Localisation and Mapping is transforming how AVs navigate dynamic environments. These systems dynamically adjust high-definition maps based on real-time road conditions. This, in turn, ensures that AVs can operate efficiently even when confronted with unexpected roadwork, new obstacles or shifting lane structures.
As AVs expand into more unpredictable and complex driving conditions, multimodal perception is emerging as a critical innovation. By integrating LiDAR, radar, thermal imaging and event-based cameras, vehicles can achieve a more comprehensive and adaptive understanding of their surroundings. This is particularly valuable for nighttime driving and adverse weather conditions (Ma et al., 2020[14]). The CCAM Partnership also prioritises the testing of such perception and localisation technologies in diverse ODDs, ensuring that automated driving functions are robust across varying European road, weather and traffic conditions (European Commission, 2024[22]).
While advances in AI-based perception and multimodal sensor fusion enable AVs to adapt to complex and adverse conditions, these technologies fundamentally depend on the integrity and readability of the physical infrastructure they navigate. Clear road markings, well-maintained signage and structurally sound roadways are prerequisites for reliable AI perception and localisation. To ensure AVs can safely interpret and interact with their environment, the underlying infrastructure must be regularly inspected and maintained. AI-powered inspection solutions, such as ASIMOB’s Autonomous Road Inspector and Enlite4’s Detekt, are helping to address this challenge (Box 5.1).
ASIMOB: Autonomous Road Inspector (Spain)
ASIMOB, a technology start-up from Bilbao, Spain, has developed an Autonomous Road Inspector, i.e. an AI-enabled tool that automates the inspection and monitoring of road infrastructure. Mounted on service vehicles or dedicated inspection units, the system combines cameras, sensors and machine- learning algorithms to detect and classify issues such as faded lane markings, damaged signage, potholes and cracks in real time. The collected data are geotagged and integrated into cloud-based dashboards, allowing road operators to prioritise maintenance tasks and monitor infrastructure degradation over time.
By providing continuous, objective and scalable assessments of road conditions, solutions like ASIMOB can support the safe deployment of AVs by ensuring that critical infrastructure elements – on which AI perception systems depend – remain visible, accurate and up to standard. ASIMOB’s platform has already been piloted in several Spanish cities and is expanding its services to support municipalities in implementing data-driven road maintenance strategies.
enlite: Mobile mapping and AI-based road asset analytics (Austria)
Detekt is an AI-powered data platform that leverages mobile mapping data to extract, catalogue and analyse road infrastructure assets. Developed by enlite, a European start-up, Detekt focuses on key components such as road damages, signage, markings and surface types. Across the European Union, 15 municipalities and transport agencies use the system to facilitate strategic planning, optimise maintenance scheduling and enhance road safety.
Key use-cases
Detecting road damage: identifies cracks, potholes and other defects, enabling timely repairs and improved prioritisation of maintenance backlogs. Pavement condition and international roughness indexes can be provided in database or visual format for long-term infrastructure planning.
Mapping road signs: offers accurate detection and categorisation of country-specific road signs, with optical character recognition to read informational signage. The platform supports custom sign ID catalogues, aiding navigation safety and maintenance planning.
Monitoring road markings: tracks wear and tear on lane markings to schedule maintenance and support AV readiness through high-resolution, updated data on lane geometry.
Classifying surface type: automatically identifies asphalt, concrete, paving stones or tile surfaces, allowing agencies to tailor maintenance strategies and optimise resource allocation.
Source: ASIMOB (2025[24]); Autonomous road inspector, https://asimob.es/en/; Detekt (2025[25]), “AI for road asset management with mobile mapping data”, https://www.detekt.it/?utm_source=website&utm_medium=button&utm_campaign=enliteAI; enlite (2025[26]), “Bringing artificial intelligence to your organization”, https://www.enlite.ai/.
Decision making and planning
Once an AV has built a detailed representation of its surroundings, AI-driven decision-making systems determine the optimal course of action most aligned with their optimisation functions. These systems must process complex and ever-changing inputs, balancing efficiency, safety and regulatory compliance. Reinforcement learning models, including deep Q-networks and policy gradient methods, play a crucial role in training AVs to perform manoeuvres such as lane-changing, merging into traffic and adjusting speed dynamically (Atakishiyev et al., 2024[18]).
In complex urban settings, hybrid AI approaches are becoming increasingly valuable. By integrating rule-based decision frameworks with probabilistic inference models, AI can navigate uncertain scenarios more reliably. This includes interpreting ambiguous pedestrian behaviour at crossings or handling unexpected driver actions at intersections (Atakishiyev et al., 2024[18]). A growing area of development is imitation learning, where AI models learn from expert human drivers to refine decision-making strategies. This approach helps AVs better replicate nuanced driving behaviours, such as smoothly yielding in dense traffic or performing subtle speed adjustments based on human driver tendencies (Hu et al., 2022[27]).
Another essential aspect of decision making involves behavioural prediction models, which analyse the actions of vulnerable road users, such as cyclists and pedestrians. By understanding subtle cues such as a pedestrian hesitating at a crossing or a cyclist shifting their weight before a turn, AI can anticipate movements and adjust accordingly to prevent collisions (Nascimento et al., 2019[28]). However, unlike human drivers, who draw on a broad spectrum of contextual and social cues (e.g. eye contact, body language and ambient traffic dynamics), AI models are limited to the visible and sensor-detectable behaviour of other road users. This can constrain prediction accuracy in complex environments, notably considering the diversity of road users (including children, older persons or persons with disabilities) (ITF, 2024[29]).
Additionally, adaptive AI control strategies continuously refine decision-making models through real-world interactions, ensuring AVs can dynamically adjust their behaviour to previously unseen traffic scenarios (Zhou et al., 2024[30]).
Motion control and safety
Translating AI-driven decisions into real-world vehicle movement requires sophisticated motion control mechanisms that prioritise precision, adaptability and safety. Model predictive control plays a key role in optimising vehicle trajectories in real time, continuously adjusting speed, acceleration and steering angles based on the surrounding environment (Garikapati and Shetiya, 2024[19]). Meanwhile, deep neural controllers allow AVs to respond dynamically to road conditions, ensuring that vehicles maintain stability even on icy roads or when encountering sudden pedestrian crossings (Atakishiyev et al., 2024[18]).
As AVs move towards full-scale deployment in dense urban environments, co‑operative manoeuvring is becoming a focal area of research. AI is being tested in multi-agent co‑ordination, where multiple AVs communicate and collaborate to optimise traffic flow. Such an approach could significantly reduce congestion and improve safety in high-density areas (Ma et al., 2020[14]). This capability is especially important at intersections, where AVs could use real-time vehicle-to-vehicle (V2V) communication to determine right-of-way, potentially reducing the reliance on traditional traffic signals in some situations.
Emerging applications of AI in automated driving reported in literature
As AV technology advances, its deployment is expected to extend beyond controlled environments and specialised pilot projects into broader applications across European mobility networks. Much research focuses on improving AV perception, decision making and motion control. However, ongoing developments are shaping how these vehicles will be integrated into urban transport systems, emergency response operations and co‑operative traffic management. The following sections highlight key areas where AVs are expected to have significant impact, supported by emerging research and technological advancements.
End-to-end AI
End-to-end AI in AVs refers to an architectural approach in which deep-learning models are trained to handle the entire driving task, directly mapping raw sensor inputs (e.g. camera, LiDAR, radar) to control actions such as steering, braking and acceleration. Unlike traditional modular pipelines that separate perception, planning and control, end-to-end systems learn to perform these tasks jointly, potentially increasing adaptability in complex or previously unseen environments.
Recent research highlights several key advantages and challenges. Chib and Singh (2023[31]) underscore that end-to-end AI allows for the emergence of more nuanced decision-making behaviours, particularly in dense urban scenarios where rule-based systems often fail. They also point out this approach reduces system complexity and latency by avoiding hand-crafted feature extraction and intermediate representations. Meanwhile, Chen et al. (2024[32]) note that, although end-to-end AI has shown promising results in simulation environments, real-world deployment still faces substantial obstacles. These include limited interpretability; safety certification concerns; and the need for extensive, high-quality labelled datasets to train reliable models. AI is also increasingly used to test and validate the reliability of automated functions, leveraging large-scale data for module training and in-vehicle decision-making testing. This further underscores the importance of hybrid architectures that retain some intermediate modular oversight for critical safety functions.
Taken together, these findings suggest the need for technological advances to resolve key issues for end-to-end AI. While end-to-end AI could simplify system integration and enhance adaptability, its deployment in safety-critical contexts like automated driving will likely depend on advances in explainability, data efficiency and regulatory alignment (Chib and Singh, 2023[31]; Chen et al., 2024[32]).
AI-guided emergency response vehicles
AI-guided emergency response vehicles are a promising deployment scenario where AVs could enhance emergency response times and operational safety. AI-powered route optimisation systems process real‑time traffic data, road conditions and emergency call locations to dynamically adjust routes, reducing delays in congested urban environments (Bram-Larbi et al., 2020[33]). Automated emergency vehicles equipped with reinforcement learning models can anticipate traffic bottlenecks and co‑ordinate with connected infrastructure, such as AI-enhanced traffic signals, to clear paths more efficiently (Rigas, Billis and Bamidis, 2022[34]). Collision avoidance mechanisms powered by LiDAR, radar and predictive analytics help ensure safe navigation through unpredictable traffic conditions. Meanwhile, augmented reality-based driver-assistance systems overlay navigation cues to support human responders in semi-autonomous scenarios (Bram-Larbi et al., 2020[33]). As these technologies evolve, automated emergency response vehicles could complement services by reducing human workload and improving response reliability.
AI-enhanced connected and co‑operative AVs
AI-enhanced connected and co‑operative AVs represent another emerging area with high potential in Europe, particularly in urban mobility and freight transport. V2V and vehicle-to-infrastructure (V2I) communication enable AVs to co‑ordinate movements in real time, reducing congestion and enhancing road safety (Garikapati and Shetiya, 2024[19]). Co‑operative AI algorithms facilitate real-time decision making for platooning, allowing AVs to dynamically form convoys that minimise aerodynamic drag, reduce fuel consumption and improve overall traffic flow (Mishra and Das, 2019[35]; Muhammad et al., 2021[17]). In dense urban environments, co‑operative AVs can interact with smart intersections, adjusting speed to avoid unnecessary stops and optimising signal phasing to improve throughput (Ma et al., 2020[14]). As many European cities implement traffic circulation plans that restrict access to certain zones based on vehicle type or time of day, AI-enhanced connected vehicles and infrastructure can help fluidify these schemes by enabling dynamic routing and supporting real-time enforcement through data sharing and intelligent control systems.
AI-enhanced AV security and cyber resilience
AI-enhanced AV security and cyber resilience is becoming an increasingly critical area as AVs become more interconnected. The reliance on AI-driven decision making, cloud-based updates and V2X communication introduces cybersecurity vulnerabilities that must be addressed before widespread deployment. Recent research highlights the role of AI in detecting and mitigating cyber threats in real time, using anomaly detection models to identify irregular sensor inputs and prevent adversarial attacks. AI‑powered encryption protocols and secure federated learning architectures are being explored to protect AV data while maintaining system efficiency (Namburi et al., 2024[36]; Onur et al., 2024[37]).
As European regulations continue to evolve, AI-driven cybersecurity solutions will play a key role in ensuring the resilience of AV networks. In the European context, the EU AI Act introduces a risk-based regulatory framework that classifies certain AI systems – such as those used in vehicle operation and safety-critical functions – as “high-risk” (see Select key EU legislative acts applying to AI in the mobility sector). This classification implies mandatory requirements for risk management, transparency, human oversight and cybersecurity. As regulations continue to evolve, AI-driven cybersecurity solutions will play a key role in protecting AV networks from hacking and system failures that could compromise passenger safety and public trust.
These deployment scenarios illustrate how AI-driven AVs are expected to move beyond experimental testing towards real-world integration. Whether in emergency response, co‑operative mobility or cybersecurity, AI will be central to ensuring that AVs operate safely, efficiently and in alignment with European transport priorities. While full-scale deployment remains a long-term objective, ongoing technological advancements are gradually paving the way towards a future in which automated driving could represent both a technical achievement and a practical contribution to enhanced mobility across the continent. Initiatives such as the CCAM Partnership (European Commission, 2024[22]) and the forthcoming European Coordinated Automated Vehicle Architecture (ECAVA) will play a central role in translating these advancements into operational mobility solutions, complementing ongoing research efforts and regulatory developments.
Insights from workshops
Most commonly discussed use-cases
Participants highlighted a range of technical domains where AI plays a critical role in enabling automated driving systems. Key use-cases include combining data from multiple sensors to understand the vehicle's surroundings; predicting how other road users will behave; creating digital replicas of real-world driving situations (digital twins); and helping vehicles make decisions by communicating with each other and with road infrastructure (V2X communication).
Pilot projects such as PoDIUM (2025[38]) and 5GMOBIX (2025[39]), presented at the ERTICO workshop, illustrated how AI can be applied to roadside infrastructure to support connected and automated driving.1 These systems use AI to process data from LiDAR and camera sensors in real time to generate co‑operative perception messages that are shared with nearby vehicles. This allows vehicles to perceive road users or obstacles outside their immediate line of sight, especially critical at intersections or in occluded urban environments. In so doing, it significantly enhances situational awareness and safety (ERTICO, 2025[40]).
Complementing these insights, a poll during the European Commission workshop shed light on the AI application areas seen as having the greatest potential for EU-wide scalability. The most frequently cited were AI tools for increased efficiency in software engineering (56%) and virtualised simulation and training (54%). Other high-potential areas included in-vehicle scenario simulation for automated driving (47%) and V2X models for multimodal traffic and mobility optimisation (42%). In contrast, AI models for infotainment, energy management and sustainability received lower scores (10‑15%), suggesting these domains may require more targeted or context-specific strategies for adoption (European Commission, 2025[41]).
At the in-vehicle level, manufacturers such as Renault, Volkswagen, BMW and Mercedes-Benz are leveraging AI to enhance automated driving capabilities and user interaction.
Renault presented a suite of AI applications supporting scene understanding, intention prediction and trajectory planning. Its “humanised technology” approach illustrates how AI can personalise interactions by adapting interfaces, anticipating user needs and learning driver behaviour over time. This, in turn, contributes to greater user trust in automated systems. Similarly, augmented reality overlays developed by Basemark provide an alternative to traditional infotainment systems. This helps drivers stay aware of their surroundings and reduces cognitive load (European Commission, 2025[41]).
Volkswagen showcased HI-Drive, a major EU-funded project co‑ordinated by Volkswagen Group Innovation. It aims to advance vehicle automation beyond SAE Level 3 towards higher levels of autonomy. Bringing together 40 partners from 13 countries, the initiative addresses key challenges around ODD fragmentation by testing high-automation functions across diverse and demanding traffic scenarios. It combines on-road trials – from motorways to urban and cross-border environments – with technology enablers such as improved perception, communication beyond line of sight and user-centred design. HI-Drive also investigates user acceptance, comfort and interaction with automated systems, providing a comprehensive basis for evaluating real-world impact and future deployment (European Commission, 2025[41]).
BMW highlighted the role of data sharing as a foundation for AI-enabled applications across the automotive value chain. It emphasises developing decentralised, trusted data-sharing mechanisms that ensure data privacy, cybersecurity and customer consent. AI use-cases showcased by BMW include PdM, in-vehicle digital services (such as app stores) and tailored usage-based insurance models. Additional applications span sustainability reporting, traffic condition monitoring and compliance with battery and supply chain regulations (European Commission, 2025[41]).
Mercedes-Benz provided insights into the large-scale deployment of connected vehicle data to support AI-based services. With over 14 million connected vehicles generating more than 25 GB of data per vehicle per hour, Mercedes-Benz has developed tools such as the City Dashboard and Developer API to enable data access for both internal and third-party applications. These initiatives facilitate a broad range of AI-supported services, including real-time traffic analytics, user-facing mobility features and interoperability with platforms such as Mobility Data Space and Catena-X (European Commission, 2025[41]).
From an industry co‑ordination perspective, the German automotive association (Verband der Automobilindustrie, VDA) introduced the ADAXO framework as a model for secure and equitable data sharing. Designed to align with EU data governance initiatives, ADAXO seeks to establish a common standard for access, integrity and transparency of vehicle data. VDA representatives underscored the need for regulatory clarity under the Data Act. It stressed that effective implementation, particularly regarding roles, interfaces and cross-border data access, will be critical to ensuring the scalability and inclusiveness of AI applications in mobility systems (European Commission, 2025[41]).
To fully unlock the potential of AI in mobility, access to high-quality, interoperable and real-time data is essential. The common European mobility data space (EMDS), as outlined in European Commission (2023[42]), is a flagship initiative to address the persistent fragmentation of transport and mobility data across the European Union.
Rather than creating a central database, the EMDS will connect and integrate existing transport data ecosystems, enabling data sharing across public and private actors in a secure, trusted and interoperable framework. It builds on cross-sectoral EU data legislation (e.g. Data Act, Data Governance Act), access points (e.g. National Access Points under the ITS Directive) and expert bodies such as the Digital Transport and Logistics Forum.
The EMDS will support a wide range of use-cases, ranging from multimodal journey planning and traffic optimisation to fleet decarbonisation and AI-enabled logistics.
Two actions supported the preparation of the EMDS:
a preparatory action (PrepDSpace4Mobility) to map ecosystems and recommend common building blocks and a reference architecture, under the Digital Europe Programme (DIGITAL) (European Commission, 2023[43])
a technical assistance study on governance and infrastructure, under the Connecting Europe Facility (Scholliers et al., 2023[44]).
An ongoing deployment project (2023‑2026) is focused on traffic and urban mobility data sharing, supporting 16 use-cases in nine cities and regions, co-funded under DIGITAL (European Commission, 2025[45]). An additional action is planned for 2025 to implement a multi-country project on mobility data infrastructure and services.
By building a federated, stakeholder-driven data space, the EMDS aims to accelerate innovation; strengthen cross-border transport integration; and enable smarter, greener and more efficient mobility across the European Union.
Source: European Commission (2025[46]), “Creating a common European mobility data space”, https://transport.ec.europa.eu/transport-themes/smart-mobility/creating-common-european-mobility-data-space_en.
Key barriers and challenges
Stakeholders consistently highlighted several systemic barriers to the deployment of AVs:
Regulatory complexity: stakeholders warned that developers and deployers must navigate coexisting and evolving frameworks (e.g. EU AI Act, GDPR, the ITS and EU Type Approval) and a range of requirements that can sometimes overlap or differ in scope, highlighting the need for clearer alignment and more harmonised guidance. This regulatory complexity is especially burdensome for start-ups and SMEs and risks slowing down AV pilots and cross-border scaling. In the context of AI-enabled tools to monitor infrastructure, the lack of a unified EU strategy for digital twins and geospatial infrastructure was highlighted as a factor increasing fragmentation and limiting the interoperability and scalability of AI systems across jurisdictions.
Data access and quality: stakeholders emphasised that high-fidelity, representative datasets, particularly those capturing rare edge cases and diverse road environments, are essential for developing safe and generalisable AI models. However, concerns persist around data protection, user consent and the lack of common European standards for sharing safety-critical data. Industry representatives highlighted the importance of purpose-specific, sovereign data-sharing mechanisms supported by robust governance models and built on customer trust. The absence of open access to municipal infrastructure data (e.g. road signs, markings and surfaces) further constrains AI innovation in road asset management. Participants saw potential in creating a European Automotive Data Platform to foster co‑operation and innovation across the industry. The value of a structured regulatory dialogue with actors across the automotive value chain was also emphasised.
Infrastructure bottlenecks and limited availability of automotive-grade, Europe-based computing resources: stakeholders noted bottlenecks to scale AI models, especially compute capacity for model training and inference. They highlighted the need for robust and layered connectivity infrastructure. This needs to span short-range V2X, terrestrial 4G/5G networks, fixed access networks (e.g. fibre) and global Internet tiers to support real-time communication, AI model deployment and continuous updates. Additionally, initiatives such as HammerHAI (2025[47]) can address compute needs at the training stage. The emergence of 3C Networks (Connected, Collaborative, Computing), combining local, regional and cloud-based compute layers, was also discussed as a pathway to enable distributed intelligence in vehicles. However, gaps in edge infrastructure, latency-sensitive data exchange and backhaul capabilities were noted as potential roadblocks. To address these, initiatives such as Axelera AI (2025[48]) were presented as strategies to promote AI compute and secure European sovereignty in model training and lifecycle management.
Outlook/other AI-based solutions
Participants highlighted the role of AI factories in providing cutting-edge, AI-optimised supercomputing services, discussing several emerging technologies. In-memory computing, for example, brings computation closer to memory to reduce latency and energy use. Meanwhile, RISC-V architectures promise to facilitate more efficient, scalable AI computing across vehicle platforms. In parallel, they emphasised the need for robust connectivity infrastructure to support real-time model deployment. Stakeholders also underscored the importance of V2X communication technologies, which allow vehicles to interact with each other and with surrounding infrastructure.
To realise these benefits at scale, substantial expansion of telecommunications infrastructure is required, particularly through the deployment of 5G standalone (5GSA) networks and advanced edge- computing capabilities. Workshop discussions indicated that AI in AVs is likely to continue progressing through focused, step-by-step advancements in specific applications, rather than through rapid, sector-wide deployment. Stakeholders noted that while technical progress in perception, behaviour prediction and co‑operative decision making is accelerating, large-scale use on public roads remains a longer-term prospect. Most near-term implementations are expected to stay limited to controlled or well-defined environments, such as geofenced urban areas, logistics hubs or industrial sites.
Promising areas for further development included the expansion of AI-powered co‑operative perception, enabling vehicles to benefit from shared data from roadside sensors; the integration of digital twins and simulation environments to support testing and validation; and improvements in human-machine interfaces that personalise interactions and foster user trust. Some participants also highlighted experimentation with AI-enabled traffic co‑ordination in mixed traffic environments, although they cautioned that scaling such solutions will require advances in data integration and infrastructure readiness.
Overall, stakeholders described the future of AI in AVs as one of gradual, targeted progress. While they expressed optimism about the potential of AI, participants stressed that technical innovation alone will be insufficient. Achieving widespread deployment will require parallel improvements in regulation, data sharing and physical infrastructure.
Key recommendations
Stakeholders expressed a broad consensus that Europe’s success in AV-related AI will depend on strong co‑ordination between public and private actors, underpinned by targeted action in five key areas:
Data
Advance implementation and strengthen stakeholder engagement within the EMDS: support EMDS implementation through targeted data-sharing incentives for priority use-cases for AV. This includes pooling and sharing high-value datasets, such as traffic, traffic accidents and incidents, naturalistic driving data, road condition, passenger flows, vehicle usage and logistics data to support AI applications. It also includes addressing concerns such as data ownership, antitrust concerns and fears of competitive disadvantage. This could be achieved by promoting secure and privacy-preserving data-sharing protocols, such as anonymisation techniques and multiparty computation, to encourage stakeholder participation.
Infrastructure and connectivity
Expand and modernise AV-supportive infrastructure: increase investment in digital infrastructure and V2X systems to enable connected and automated driving. Promote V2X cloud platforms with ultra-low latency and built-in cybersecurity features for real-time, secure data exchange between vehicles and infrastructure. Standardise V2I and V2V communication protocols at the EU level, including minimum requirements for encryption, authentication and threat detection.
Improve connectivity networks for mission-critical AV operations: accelerate the rollout of 5G networks and ensure ubiquitous, high-quality coverage across Europe to guarantee quality of service for AV functions.
Fund infrastructure upgrades: create dedicated funding instruments for infrastructure retrofitting and AV-supportive road environments.
Promote digital twin strategies for infrastructure, including within the EMDS: encourage the creation of an EU-wide framework for digital twins of road infrastructure to standardise data formats, support PdM and enable interoperability across cities and regions. Harmonised digital twin strategies can improve asset management and facilitate integration of AV systems with real-time infrastructure data.
Enable drone-based infrastructure monitoring: harmonise regulations for unmanned aerial vehicles (UAV) (drones) across Member States to allow controlled use of drones for capturing road conditions and infrastructure data, reducing costs and increasing coverage for digital twin creation.
AI solutions, software and interoperability
Promote open and interoperable AI ecosystems: encourage open-source and interoperable solutions by promoting open AI algorithms, model weights and standards to reduce dependency on dominant players and stimulate innovation. Enable interoperability across different vehicle platforms, suppliers and nations to build a unified AV ecosystem.
Support deployment of sector-specific AI tools: raise awareness among public authorities and operators about the availability of cost-effective AI tools for infrastructure, and support pilot projects that showcase their value. Streamline procurement processes to favour AI-enabled services over hardware-heavy investments when feasible.
Regulatory frameworks and testing environments
Enable flexible, harmonised AV testing and regulation: enhance regulatory frameworks to enable innovation by simplifying and clarifying the EU AI Act and related automotive AI regulations to make them understandable and actionable for developers. Promote co‑ordinated regulatory experimentation, involving Member States, local actors and standardisation bodies. Support cross-border test beds and large-scale pilots to harmonise standards and accelerate the validation of AV technologies.
Ensure coherence in infrastructure procurement rules: promote shared procurement platforms and regional pooling of resources to increase efficiency and sustainability.
Skills, talent and collaboration
Develop AI talent and foster collaboration: support EU-wide training and upskilling initiatives in automotive AI and system integration to build a skilled workforce.
Strengthen and pursue public-private partnerships: build on initiatives such as the CCAM Partnership and the planned ECAVA to foster collaboration between academia, start‑ups, OEMs and infrastructure operators. Support expansion of such partnerships where relevant, and consider further stakeholder consultation to identify remaining gaps in co‑ordination, knowledge transfer or pilot deployment.
Create shared AI innovation hubs: establish a “CERN for AI” to promote open, trustworthy and mission-driven innovation at the European level.
AI for public transport systems
AI is becoming increasingly important in modern public transport, driving efficiency, resilience and an improved passenger experience. As urban populations grow and mobility patterns shift, European cities are increasingly leveraging AI to manage transport networks in a more flexible and responsive manner. AI enables real-time demand forecasting, optimises fleet deployment and facilitates multimodal integration. In so doing, it ensures seamless connectivity between buses, trains, micromobility services and ride-hailing platforms (Nikitas et al., 2020[49]).
In addition to operational improvements, AI enhances PdM, reducing disruptions by detecting infrastructure wear and mechanical failures before they occur (Ding et al., 2023[50]). AI-powered security systems strengthen passenger safety, monitoring real-time video feeds for potential threats, while identifying fare evasion patterns to minimise revenue loss (Adanyin and Odede, 2024[51]; Jevinger et al., 2024[52]). While many of these technologies have been implemented across European transport networks, the continuous advancements in AI are expanding the scope of its applications.
Main use-cases reported in literature
AI for network optimisation and service efficiency
Public transport networks must balance efficiency with reliability, ensuring that services meet fluctuating demand while maintaining cost effectiveness. AI-driven predictive analytics are transforming transit planning by allowing agencies to anticipate ridership patterns and adjust schedules dynamically. For instance, in Sofia (Bulgaria), the Urban Mobility Centre partnered with a technology provider to use AI with real-time images from onboard cameras to classify bus occupancy levels (UITP, 2025[53]). Machine learning (ML) models process historical and real-time occupancy data alongside external factors such as weather and traffic conditions, optimising fleet allocation accordingly (Cohen, 2024[54]). Long Short-Term Memory (LSTM) and Bidirectional LSTM models have demonstrated strong performance in short-term demand forecasting using passenger smart card and sensor data (Liyanage et al., 2022[55]). These AI‑powered scheduling systems help prevent overcrowding, reduce wait times and increase the efficiency of resource allocation (Paiva et al., 2021[56]). The National Transport Authority of Ireland deployed a ML prediction engine via the Trapeze platform to enhance arrival time accuracy. By dynamically adjusting the weighting of short-term and historical data and integrating traffic, weather and passenger load information, the system improved prediction accuracy by 13%, with full deployment expected in 2025 (UITP, 2025[53]).
Beyond scheduling, AI is reshaping fleet management by dynamically adjusting vehicle distribution based on evolving passenger needs. AI-powered optimisation models help transport agencies operate mixed fleets of electric, hybrid and conventional vehicles more effectively, ensuring energy-efficient operations (Abirami et al., 2024[57]). As more European cities transition towards electrification, AI-driven solutions are also integrating real-time energy management, co‑ordinating vehicle charging with grid demand to prevent power shortages.
AI is also improving transit co‑ordination at a city-wide level. Intelligent traffic management systems use reinforcement learning algorithms to adjust traffic light sequencing dynamically, prioritising buses and trams at intersections to reduce delays (Patil et al., 2024[58]). These AI-driven traffic signal priority systems are critical for multimodal integration, improving the reliability of public transport services while reducing congestion across the broader transport network. Examples from Hamburg, Barcelona and pan-European initiatives are highlighted below.
In Hamburg, the State Office for Roads, Bridges and Waterways and the Planung Transport Verkehr GmbH (PTV) partnered in the “Transmove” project to enhance smart mobility forecasting. As part of this initiative, PTV developed the AI-based “PTV Optima” system, which integrates an agent-based traffic model with real-time data to generate short-term traffic forecasts. Delivering predictions for a 5‑30 minute period, the system enables operators and traffic control centres to make proactive, data-driven decisions, supporting dynamic traffic management and improving public transport flow across the city (PTV Group, 2024[59]; UITP, 2025[53]).
The city of Barcelona also uses AI-powered adaptive traffic signal systems, which respond to real-time traffic data. These systems adjust signal timings dynamically to reduce congestion and prioritise public transport vehicles at intersections. This, in turn, enhances overall traffic flow and supports modal shift towards sustainable transport.
In a third example, the Artificial Intelligence and Mobility Operations project combines AI-powered traffic management with a foundation model trained on decentralised mobility data from across Europe (Box 5.3).
Initiatives such as the Artificial Intelligence and Mobility Operations (AIAMO) project illustrate how AI foundation models can be used to unify fragmented traffic and transport data across jurisdictions.
While Intelligent Transport Systems (ITS) do not embody intelligence as such, they are advanced applications. They aim to provide innovative services relating to different modes of transport and traffic management. In so doing, they enable various users to be better informed and make safer, more co‑ordinated and “smarter” use of transport networks (Official Journal of the European Union, 2010[60]). ITS does not involve vehicle manufacturing but focuses on operational management of transport infrastructure. The ITS community is organised globally, with regional branches in Europe (via the ERTICO network), Asia-Pacific and the Americas.
Within this framework, the AIAMO project represents a flagship initiative aimed at advancing AI-based traffic and mobility management in Europe. Funded by the German Federal Ministry of Digital and Transport, AIAMO is co‑ordinated by ITS Germany e.V. and brings together 13 partners from the mobility, environment and AI sectors. The project, which builds on the country’s mobility data space initiative, is implemented in collaboration with ITS industry stakeholders, such as Swarco and Yunex.
AIAMO focuses on integrating and training large-scale mobility data using a foundation model architecture rather than developing final user-facing applications. At its core is the AIAMO Nexus, a model designed to train and process large volumes of mobility and environmental data. AIAMO Nexus functions as a shared integration layer that aggregates, semantically validates and processes data from diverse sources, including traffic signals, public transport systems, environmental sensors and national mobility data spaces. The resulting AI-ready datasets are then made available to public and private actors for use in their own applications. Therefore, AIAMO itself does not produce end-user tools; rather, it provides data-ready outputs that municipalities and industry actors can use to develop AI-enabled services.
The project is designed to support environmentally sensitive, demand-responsive mobility management, particularly in urban settings. AI applications informed by AIAMO Nexus include real-time traffic signal optimisation, multimodal co‑ordination and public transport prioritisation. As its core principle, data infrastructure (e.g. national access points) should remain decentralised and locally governed. Conversely, the training of foundation models should be centralised to ensure consistency, scale and generalisability.
The project emphasises data sovereignty and interoperability, ensuring that data from various jurisdictions can be used while remaining under the control of local authorities. It also seeks to lower barriers for SMEs and smaller cities to engage with AI systems by promoting modular access to mobility data and AI outputs.
Pilots in Leipzig and London set to start in July 2025 will test the system’s ability to support AI-assisted traffic signal optimisation, public transport prioritisation and emissions-sensitive routing. AIAMO also includes work packages on digital twins, real-time traffic and environmental modelling and multimodal mobility integration.
The AIAMO model is considered replicable across other data-rich, regulated sectors such as energy, agriculture and health, where foundation models could be applied to large-scale operational data.
Source: AIAMO (2025[61]), “AIAMO – Artificial intelligence and mobility”, https://www.aiamo.de/en.
AI for operational resilience and maintenance
Public transport systems rely on infrastructure and rolling stock that must be continuously maintained to prevent service disruptions. AI is revolutionising maintenance strategies by enabling predictive monitoring of vehicles, tracks and facilities. AI-driven big data platforms continuously analyse sensor data to identify early signs of mechanical wear, helping transport agencies schedule preventive maintenance and avoid costly failures (Güven and Sahin, 2022[62]). AI-powered anomaly detection further enhances system reliability by flagging deviations in vehicle performance that could indicate potential breakdowns (Jevinger et al., 2024[52]).
Infrastructure monitoring has also seen significant advancements with AI-driven analysis of structural health. Technologies such as UAVs equipped with deep-learning models are detecting track misalignment; structural weaknesses in bridges; and cracks in tunnels, allowing for proactive interventions (Bianchi et al., 2025[63]). AI-based energy management systems further contribute to operational resilience, helping cities optimise the charging and power distribution of electric buses, trams and metro systems (Patil et al., 2024[58]). By improving the efficiency of maintenance strategies and reducing unplanned downtime, AI is enhancing both cost effectiveness and service reliability in public transport.
AI for passenger experience and security
Public transport efficiency is ultimately measured by passenger experience, and AI is playing a growing role in improving comfort, accessibility and security.
AI-driven real-time passenger flow management systems analyse video feeds and smart card transactions to detect overcrowding. This enables transport operators to respond dynamically by deploying additional vehicles or modifying service frequencies. AI-powered movement pattern analysis also helps predict congestion, particularly in metro stations and other high-traffic hubs (Jevinger et al., 2024[52]).
AI-enhanced security solutions can improve passenger safety through automated video surveillance and anomaly detection. Using movement tracking, AI-based systems can continuously monitor transit environments for suspicious activities to detect potential threats in real time (Cohen, 2024[54]). Fare fraud detection is another area where AI shows great potential, with automated transaction monitoring helping identify irregular payments and prevent revenue loss (Adanyin and Odede, 2024[51]; Jevinger et al., 2024[52]). For instance, in Barcelona, the Railways of the Government of Catalonia (Ferrocarrils de la Generalitat de Catalunya, FGC) has deployed the AWAAIT AI system since 2015 to detect fare evasion at urban stations. The system analyses real-time video footage at fare gates and alerts ticket inspectors via a mobile app, allowing for targeted interventions without inconveniencing law-abiding passengers (UITP, 2025[53]). According to FGC, the system led to a 70% reduction in fare evasion during pilot tests (AWAAIT, 2025[64]). These AI-driven security enhancements may contribute to a more reliable and trusted public transport system, ensuring that transit networks remain both safe and financially sustainable.
Recent developments have also brought the integration of LLMs into public transport applications, where they support both staff operations and customer service. By leveraging natural language understanding and generation, these models help streamline communication, reduce response times and enhance user assistance. For instance, in Italy, Club Italia developed “Velvet,” an open-source chatbot powered by LLMs. It provides users with real-time service information, including schedule updates and weather-related travel advice. Velvet is being piloted in collaboration with partners such as IntercentER, UniMarconi and the Veneto Region, supporting public transport operators in Italy’s central and southern regions, especially in Sicily (UITP, 2025[53]).
Emerging applications of AI in public transport reported in literature
As AI technologies continue to evolve, their applications in public transport are extending beyond deployments, supporting more adaptive, integrated and sustainable mobility solutions. European cities are at the forefront of this expansion, using AI to bridge gaps in multimodal transport, facilitate on-demand mobility and enhance sustainability. These applications are not entirely new but represent an evolution of capabilities, allowing AI-driven solutions to refine transport planning and management.
AI for multimodal integration and dynamic route optimisation
AI-driven decision making is revolutionising multimodal transport integration, a key objective for cities seeking to provide seamless mobility. Traditionally, public transport systems have relied on static timetables and predetermined routes, making it difficult to adapt to real-time fluctuations in demand. AI‑enhanced MaaS platforms now enable real-time multimodal co‑ordination by integrating buses, trains, micromobility services and ride-hailing platforms into a single optimised network (Nikitas et al., 2020[49]).
AI-powered dynamic route optimisation models improve passenger experience by continuously adjusting services based on congestion levels, weather conditions and user demand (Cohen, 2024[54]). Reinforcement learning algorithms analyse vast datasets from transport networks enabled by the Internet of Things (IoT), allowing systems to propose the most efficient routes dynamically (Chu et al., 2024[65]). These capabilities extend beyond traditional public transport management, offering predictive congestion avoidance and AI-guided journey planning that optimises transfers between different transport modes. By bridging previously fragmented transport networks, AI has great potential to reshape urban mobility into a truly interconnected system.
A practical example comes from Barcelona, where the deployEMDS pilot has established a multi-operator data space ecosystem to enhance integration between bus fleets and on-demand transport (deployEMDS, 2025[66]). Led by ATM Barcelona in partnership with Nommon, i2CAT and Transports Metropolitans de Barcelona (TBM), the initiative uses AI-powered predictive analytics to improve planning and co‑ordination between regular lines, on-demand services and demand-responsive transport. The pilot demonstrates how AI and shared data infrastructures can support more adaptive, efficient and user-centred transit systems.
Automated last-mile shuttles and on-demand transit
While fixed-route buses, trams, regional rail and metro systems remain the backbone of public transport, AI is enabling a shift towards more flexible, demand-responsive services. AI-driven on-demand transit platforms leverage predictive analytics to anticipate travel patterns, dynamically adjusting routes and schedules based on real-time ridership data (Liyanage et al., 2022[55]). These services are particularly beneficial in low-density areas where traditional public transport options are costly and underused.
Automated shuttles represent another emerging application of AI in public transport. These self-driving vehicles are being piloted in controlled environments such as university campuses, business districts and residential areas, where they provide last-mile connectivity to major transit hubs (Tarkiainen et al., 2021[67]). AI-powered fleet management systems monitor vehicle availability and passenger demand, dynamically dispatching automated shuttles to optimise service efficiency (Turno and Yatskiv, 2023[68]). V2I communication further enhances operational performance by enabling AI systems to interact with traffic signals and road sensors, ensuring safer and more efficient operations (Mirindi, 2024[69]).
A noteworthy example is the EU-funded SHared automation Operating models for Worldwide adoption (SHOW) project, one of the largest initiatives piloting connected, co‑operative and automated mobility in Europe. Spanning 69 partners across 13 Member States, SHOW tests autonomous shuttles and AI‑enabled on-demand mobility services in over 20 urban areas, including mixed traffic environments and multimodal hubs. The project’s core objective is to demonstrate how shared, AI-powered fleets can complement traditional public transport by offering flexible, demand-responsive services that enhance last-mile connectivity, especially in underserved areas. AI plays a central role in fleet orchestration, passenger demand prediction and route optimisation, allowing autonomous shuttles to dynamically adapt to real-time traffic conditions and user needs. Furthermore, SHOW integrates V2I and vehicle-to-network communication to improve safety and traffic flow, while also evaluating user acceptance and operational viability. By embedding AI within a scalable and interoperable architecture, SHOW provides a blueprint for the broader integration of autonomous systems into urban mobility networks across the European Union (Rataj et al., 2025[70]).
Two other recent EU initiatives are notable. For its part, metaCCAZe (2025[71]) tests and demonstrates cutting-edge technologies for zero-emission and automated mobility solutions for both passengers and freight. MOBILITIES FOR EU (2025[72]) pilots innovative electrified and automated mobility technologies in diverse real-world urban settings.
As these systems mature, automated public transport is expected to play a critical role in supplementing networks, reducing operational costs and improving accessibility in underserved areas. However, regulatory and infrastructural barriers remain key challenges to large-scale deployment. This necessitates continued research, pilot programmes, investments in infrastructure and policy measures such as updating and harmonising regulations.
AI-driven sustainable and inclusive mobility
AI is increasingly being applied to enhance the sustainability and inclusivity of public transport. AI-driven environmental analytics provide transport authorities with real-time emissions tracking, enabling the optimisation of bus and train operations to minimise carbon footprints (Alqasi, Alkelanie and Alnagrat, 2024[73]). ML models assess energy consumption patterns across entire transit networks, identifying opportunities to reduce fuel usage and integrate renewable energy sources.
Projects such as the EU-funded UVAR Box demonstrate the role of digitalisation in paving the way for AI-supported environmental optimisation in urban mobility. By enabling cities to digitise Urban Vehicle Access Regulations (UVARs) (e.g. low-emission zones, congestion charges and traffic restrictions), the project lays the foundation for more dynamic and intelligent management of traffic flows. While UVAR Box itself does not develop AI tools, its digital infrastructure creates the necessary conditions for integrating AI-driven environmental analytics. This allows authorities to monitor vehicle-related emissions in real time and implement adaptive strategies to minimise pollution. Such efforts support the broader shift towards sustainable, data-informed urban mobility planning and strengthen the capacity of public transport systems to contribute to climate objectives (Rataj et al., 2025[70]).
AI is also being deployed to enhance accessibility and equity in public transport. AI-powered accessibility assessment tools analyse ridership data to identify mobility gaps and propose solutions for passengers with disabilities. For example, AI-driven route planning systems integrate real-time accessibility information, such as lift functionality at metro stations and bus stop conditions, to assist passengers with mobility impairments (Patil et al., 2024[58]).
Insights from interviews
Most commonly adopted use-cases
Interviewees said AI adoption in public transport was at an exploratory stage, with most implementations still in targeted pilots or early deployment. Rather than being fully integrated into core operations, AI solutions are often limited to supporting specific departments or isolated functions. Among the most reported use-cases were those that leverage sensor and video infrastructure to improve operational safety. For instance, interviewees mentioned systems that analyse video footage to detect platform intrusions or identify patterns of safety incidents across the network. These tools are primarily used in post-event analysis but were also seen as potentially useful for informing route design, infrastructure upgrades or staff allocation strategies.
Some operators also highlighted the role of AI in supporting fleet electrification efforts, particularly by optimising charging schedules for electric buses based on real-time demand forecasts, vehicle availability and route characteristics.
In addition, interviewees referenced trialling fraud detection tools that identify suspicious travel behaviour. While these efforts are still early stage, there was also interest in applying AI to improve passenger experience through, for example, automated chatbots or smart notification systems. These were not yet widely adopted but were viewed as promising areas for further exploration, particularly in larger urban networks.
The International Association of Public Transport assessed the maturity and adoption level of AI applications (Table 5.3), indicating that none is advanced, i.e. fully deployed across multiple departments and critical to core operations.
Notes: “Planning” refers to applications not deployed, but under consideration for future integration; “Early stage” refers to limited deployment, primarily in testing or pilot phases; “Intermediate” refers to partially deployed AI applications, supporting specific functions or departments.
Source: UITP response to interview.
UITP also highlighted regional disparities in the adoption of AI in public transport across the European Union. Western Europe is leading the way, driven by robust infrastructure and sustained public funding. Nordic countries stand out for their innovative applications, particularly those aligned with sustainability goals. In contrast, southern Europe is at an intermediate stage, showing growing interest but facing limitations in scale and consistency. Eastern Europe, meanwhile, remains in the early adoption phase, largely due to constrained resources and underdeveloped infrastructure.
A significant urban-rural divide was also noted. Urban areas benefit from higher population density, larger budgets, and stronger institutional and private sector support, enabling more rapid deployment of AI technologies. In rural regions, adoption is more limited, with a continued focus on traditional mobility challenges rather than advanced AI systems.
Key barriers and challenges
Across all interviews, stakeholders pointed to a set of recurring barriers that limit the wider deployment of AI in public transport systems.
Lack of off-the-shelf, integration-ready solutions: many AI tools on the market require substantial customisation to work within legacy IT systems or match the operational realities of public transport providers. Since many urban mobility systems rely on outdated infrastructure, integrating AI with legacy systems can be technically challenging and resource intensive. Fragmented infrastructure and outdated signalling systems were seen as limiting factors for AI implementation in smaller or older networks.
Data availability and quality: AI systems rely heavily on high-quality, real-time data, yet many cities face issues such as fragmented data ownership, variable data standards and limited interoperability between systems. Several interviewees noted that while internal operational data are often well structured, data sharing and governance within organisations can still be fragmented across departments. This challenge is further compounded when trying to connect internal data with external sources, such as weather information or crowd movement patterns. Stakeholders also emphasised the importance of data aggregation and standardisation. They noted that foundation models can only function effectively if mobility data from multiple jurisdictions can be accessed, verified and processed. The lack of semantically consistent and structured data was identified as a key technical bottleneck.
Cybersecurity risks: the growing digitalisation of transport infrastructure introduces increased vulnerability to cyberattacks. Interviewees stressed the need for robust cybersecurity frameworks to protect sensitive data and ensure system resilience.
Size and capacity of digital infrastructure: as real-time data flows continue to grow exponentially, the capacity and security of supporting digital infrastructure (e.g. IT networks, cloud storage, data pipelines) must be significantly enhanced to meet future demands. Stakeholders noted that achieving real-time AI integration will require co‑ordinated investments in IT systems, sensors and cloud connectivity, especially in under-resourced municipalities.
Organisational challenges: interviewees reported a mismatch in technical capacity between operational teams and IT or data departments, limiting collaboration and delaying implementation. Stakeholders observed that municipalities often lacked internal AI expertise and depended on external actors to interpret or operationalise model outputs. This skills gap is compounded by the broader shortage of AI and data science expertise in the public transport sector.
Cost and resource constraints: AI systems require significant upfront investment for development, deployment and maintenance. Smaller cities or under-resourced operators struggle with the financial and human capital needed to support long-term AI integration.
Concerns around liability and public trust: interviewees highlighted resistance to change and limited acceptance, especially in safety-critical domains such as video surveillance or automated decision making. One interviewee noted that even technically sound systems can face resistance from staff or unions if perceived as replacing human judgement.
Procurement hurdles: the lack of standardised procurement processes for AI solutions complicates vendor selection and increases the risk and cost of implementation. On the demand side, this is particularly problematic for operators seeking to pilot AI technologies on a limited budget. Meanwhile, companies identified this as a key issue when working across jurisdictions with differing technical standards and contracting requirements. Discussions with companies further highlighted the high complexity and duration of procurement procedures, which often exclude start-ups offering innovative AI solutions. This underscores the need to consider more flexible and simplified procurement pathways (e.g. dedicated innovation procurement procedures or tailored frameworks for small-scale, experimental solutions) to better support emerging providers.
Regulatory barriers: the lack of harmonised EU-wide regulations for AI and AVs creates uncertainties. Stakeholders stressed the need for clearer guidance, particularly on the AI Act and GDPR, to ensure compliance while fostering AI adoption. Interviewees pointed to six areas requiring greater clarity at the EU level. These comprise i) definitions and classifications under the AI Act’s risk categories; ii) certification requirements for AI systems; iii) well-defined liability frameworks in cases of system failure or accidents; iv) guidance on standardisation and interoperability across borders and institutions; v) expectations for identifying, mitigating and reporting bias in AI systems; and vi) specific guidelines for incorporating environmental considerations into AI system assessments.
Outlook/other AI-based solutions
Interviewees generally expressed cautious optimism about the future role of AI in public transport. While acknowledging that adoption is uneven and often experimental, they suggested that AI applications are likely to grow in importance as successful pilots mature and become embedded into core operations. Some promising areas mentioned included automated risk forecasting, dynamic scheduling and AI-driven passenger communication tools, such as chatbots or intelligent notification systems.
In addition, some interviewees pointed to emerging use-cases focused on inclusion and accessibility. For example, AI systems can support passengers with disabilities by improving wayfinding or customising travel information. There was also interest in collaborative innovation, with several stakeholders reporting partnerships with universities or research centres to explore AI-based solutions that go beyond off-the-shelf tools.
Interviewees also recognised public transport systems as being well positioned to support AI integration thanks to a range of strengths. Mobility networks generate rich data streams from global positioning systems (GPS), sensors, ticketing systems, surveillance and IoT infrastructure. At the same time, they maintain extensive historical data archives. Both are critical for training and optimising AI models. Many cities already possess smart infrastructure, including adaptive traffic signals and digital payment platforms, which can serve as a solid foundation for AI-enabled upgrades. In addition, public-private partnerships involving technology firms, start-ups and local authorities have created a collaborative environment conducive to innovation. With millions of daily users, transport systems provide direct public feedback loops, helping to refine and improve AI applications. Moreover, the availability of EU, regional and local grants aligned with smart mobility goals supports experimentation and scaling. Controlled environments, such as metro networks or dedicated lanes, offer ideal conditions for testing new technologies like AVs or predictive pricing. Finally, the trusted status of public transport as a socially equitable service provides a strong foundation for public acceptance of AI-driven solutions.
However, several interviewees also stressed that the success of future AI deployment would depend on aligning expectations with operational reality. In their view, vendors often overpromise on AI projects, leading to misaligned expectations and disillusionment when outcomes fall short. As a result, future deployments would need to focus on well-defined problems where value can be demonstrated incrementally.
Key recommendations
Interviewees offered several concrete recommendations to facilitate the wider and more effective use of AI in public transport systems:
Data
Foster interoperable and secure data sharing: promote EU-level data pooling and sharing mechanisms to support AI-enabled public transport applications. Facilitate the exchange of high-value datasets such as passenger flows, service disruptions, vehicle locations, operational performance, accessibility features and multimodal transfer data. Address data ownership, privacy and antitrust concerns by encouraging use of secure, privacy-preserving data-sharing protocols, including anonymisation techniques and multiparty computation. Incentivise participation across public and private operators, municipalities and MaaS providers to ensure broad and equitable access to critical mobility data.
Reinforce and scale foundational data infrastructure efforts within EU and national mobility data space infrastructure: provide funding for initiatives to further promote data standardisation, integration and interoperability across mobility data platforms. Stakeholders noted the value of a centralised foundation model trained on decentralised data sources.
Invest in data management capabilities: strengthen local capacities to aggregate, clean and standardise data from diverse systems. Encourage use of common data schemas and semantics to improve data readiness for AI processing.
Facilitate collaborative data platforms: establish data-sharing platforms to enable collaboration between transport authorities, operators and technology providers. Ensure that data spaces are accessible to smaller actors, including SMEs and municipalities.
Infrastructure and connectivity
Build AI-ready infrastructure across cities: establish common procurement frameworks (at national or EU level) to develop AI-compatible infrastructure (e.g. connected traffic signals, IoT‑enabled assets), reduce costs, increase interoperability and provide quality assurance for public operators.
Support infrastructure upgrades for smaller cities: provide targeted financial support (vouchers or grants) to help smaller cities and municipalities upgrade legacy systems and adopt infrastructure such as sensors, IoT systems and IT network infrastructure for real-time data collection and AI deployment.
Modernise legacy systems for real-time AI use: support the upgrade of legacy operational technology and IT systems and deployment of IoT-enabled infrastructure for real-time data collection.
AI solutions, software and interoperability
Promote open and interoperable AI ecosystems: encourage open-source and interoperable AI solutions to reduce vendor lock-in and enable broader adoption by smaller operators.
Enable experimentation through pilots and virtual testing: support cities and operators through pilot programmes and digital twin simulations with dedicated funding to allow cities to test and evaluate AI solutions before scaling.
Advance AI for sustainability goals: provide funding for projects that combine AI with green mobility initiatives, such as carbon reduction and air quality monitoring.
Develop shared AI infrastructure: support the creation of shared European AI infrastructure, including foundation models, open datasets, benchmarking environments and testing platforms, to enable the training, deployment and optimisation of transport-related AI applications.
Regulatory frameworks and testing environments
Clarify AI regulation in public transport: clarify how the AI Act applies to public transport settings in order to reduce complexity and support responsible experimentation.
Enable regulatory sandboxes for safe experimentation: encourage countries to establish regulatory sandboxes where cities and regions can test AI solutions in controlled environments with regulatory flexibility in early deployment phases.
Skills, talent and collaboration
Invest in skills and workforce development: invest in internal upskilling and change management, especially for non-technical staff. Create targeted initiatives for AI workforce development, including upskilling programmes, in the public transport sector.
Facilitate access to expertise and funding support: provide targeted technical assistance for grant applications and project management to improve access to EU funding. Support recruitment of AI specialists, data scientists and software engineers within public transport organisations.
Foster innovation partnerships and research: strengthen ongoing efforts to connect transport authorities with research institutions and technology providers and support applied research collaborations that tailor AI solutions to real-world operational needs. Further encourage the development of dedicated regional innovation hubs and long-term public-private-academic partnerships to promote collaboration, innovation and knowledge sharing.
Promote public trust and peer learning: fund campaigns to educate citizens about the benefits and challenges of AI in mobility, addressing privacy, safety and employment concerns. Foster peer learning and best practice sharing, enabling operators to learn from each other’s successes and failures with AI tools.
Fleet management (freight transport)
AI is transforming fleet management in freight transport, a domain that broadly includes heavy-duty road transport, rail and maritime freight operations. While heavy logistics spans multiple modes, this section focuses on the use of AI in road freight and maritime shipping. In both modes, AI technologies are increasingly embedded in fleet operations to enhance operational efficiency, safety and sustainability. AI‑driven systems support functions such as PdM, real-time route optimisation and intelligent safety monitoring, thereby improving fleet reliability, saving costs and reducing risks.
As AI continues to evolve, its role in fleet management is expanding beyond optimisation and automation. Applications are emerging in predictive supply chain resilience, AI-driven sustainability analytics and multimodal freight integration. This reflects a shift towards proactive, data-driven logistics strategies that enhance adaptability in a rapidly changing global market
Main use-cases reported in literature
AI applications in fleet management can deliver substantial improvements in logistics efficiency, asset longevity and safety. The literature highlights three key AI use-cases that are most deployed in fleet operations: i) PdM and asset management; ii) AI-driven route optimisation and scheduling, and iii) AI‑based safety monitoring and driver assistance. Each of these applications leverages ML, predictive analytics and real-time data processing to improve fleet management outcomes.
Predictive maintenance and asset management
PdM represents one of the most transformative AI applications in fleet management. By leveraging real-time data, it enables the proactive identification of potential mechanical failures, reducing unplanned downtime and lowering overall maintenance spending. Traditional fleet maintenance is based on fixed servicing schedules or reactive repairs, both of which can lead to inefficiencies, for instance, due to breakdowns. AI-driven PdM shifts the model towards condition-based monitoring, where vehicles are serviced based on actual status and performance of vehicle components rather than predetermined timelines (Woschank, Rauch and Zsifkovits, 2020[74]).
Diagnostic systems powered by AI analyse data from telematics sensors, engine control units and historical maintenance records to detect anomalies in engine performance, fuel consumption or component degradation (Durlik et al., 2024[75]). ML models identify patterns that suggest early signs of mechanical failure, allowing fleet operators to schedule repairs before failures occur. This approach improves fleet uptime and reduces the risk of unexpected breakdowns, particularly in long-haul transport where vehicle reliability is critical (Du Plessis et al., 2025[76]).
In addition to anticipating equipment failure, AI contributes to optimised inventory management by aligning spare parts procurement with PdM forecasts. This reduces unnecessary expenses associated with overstocking, while preventing costly delays caused by unavailable critical parts (Tang et al., 2022[77]).
A prominent example of AI-powered PdM comes from the maritime sector, where Maersk Line has deployed an AI-driven system across its container ship fleet. The system continuously collects and analyses real-time data from onboard sensors, which measure variables such as engine temperature, vibration and pressure. ML models detect early indicators of mechanical issues before failures occur. This approach has significantly reduced unscheduled downtime, enhanced machinery lifespan and enabled more precise maintenance scheduling. According to Maersk, these predictive capabilities have helped cut maintenance costs by up to 20%, while improving overall fleet reliability and operational efficiency (digitaldefynd, 2025[78]).
AI-driven route optimisation and logistics planning
AI-driven route optimisation improves logistics by identifying the best transport routes based on real-time traffic, weather conditions and delivery requirements. Unlike traditional planning methods that rely on static maps and fixed schedules, AI continuously analyses real-time data to optimise delivery timelines, reduce fuel consumption and avoid congestion (Olugbade et al., 2022[79]).
ML models can also assess past delivery performance, seasonal traffic patterns and fuel usage to anticipate potential delays and recommend alternative routes. Fleet operators increasingly rely on AI‑assisted dispatching systems that combine real-time GPS data, vehicle availability and customer orders to allocate vehicles and schedule deliveries (Shi, 2022[80]).
AI systems can also support more complex decision making, balancing factors such as cost, emissions, vehicle capacity and service-level agreements. For example, some AI systems prioritise the most fuel‑efficient routes, while others optimise for the shortest delivery time (Du Plessis et al., 2025[76]). The French shipping company CMA CGM, for example, has implemented AI-driven route planning to optimise vessel trajectories based on real-time environmental data, vessel characteristics and fuel efficiency metrics. According to the company, this has led to reductions both in fuel consumption and GHG emissions, while enhancing service reliability and punctuality (digitaldefynd, 2025[78]). In April 2025, CMA CGM announced a five-year strategic partnership with the French AI company Mistral AI, aiming to transform operations across the shipping, logistics and media sectors through advanced AI solutions (CGA CGM Group, 2025[81]).
AI-based safety monitoring and driver assistance
Ensuring safety in fleet operations is a top priority for the transport industry, particularly in freight transport, where vehicle size, weight and long-haul distances increase accident risks. AI-powered safety monitoring systems are transforming fleet management by continuously assessing driver behaviour, vehicle stability and road conditions. This is significantly reducing accident rates and improving compliance with safety regulations. AI-driven telematics and ADAS leverage ML, computer vision and real-time sensor data to prevent collisions, mitigate driver fatigue and enhance vehicle safety performance (Yang et al., 2024[82]).
One of the most prominent AI applications in fleet safety monitoring is real-time driver behaviour analysis. AI-powered telematics systems collect and process vast amounts of data on braking intensity, acceleration patterns, lane discipline and steering behaviour. ML models identify risk-related behaviours such as frequent hard braking, sudden lane changes or prolonged speeding. By monitoring these parameters in real time, AI can issue alerts for drivers and fleet managers, facilitating immediate corrective action (Abduljabbar et al., 2019[83]).
In addition to monitoring driver behaviour, AI is also advancing predictive accident prevention. By integrating data from radar, LiDAR, GPS and road sensor, AI can assess external conditions and anticipate potential risks from road hazards, poor weather conditions and sudden traffic congestion. These insights support predictive braking, lane-keeping assistance and collision avoidance features that automatically adjust speed, maintain safe distances and take evasive actions when necessary. These ADAS features are particularly effective in reducing accidents caused by rear-end collisions and loss of control in adverse conditions (Durlik et al., 2024[75]).
AI is also improving fleet-wide safety enforcement and performance monitoring. AI-powered dashboards provide fleet managers with detailed analytics on individual driver performance, identifying trends in unsafe driving habits over time. These systems generate safety scores, allowing operators to implement targeted driver coaching programmes and incentive-based safety compliance strategies (Yang et al., 2024[82]).
Emerging applications of AI in fleet management reported in literature
As global supply chains grow in complexity, environmental regulations become more stringent and logistics networks expand across multiple transport modes, new AI-driven capabilities are emerging to meet these evolving challenges. The literature highlights three key areas of emerging AI applications: i) AI-enabled predictive supply chain resilience; ii) AI-driven sustainability analytics and green fleet optimisation; and iii) AI in multimodal freight and automated fleet co‑ordination, which improves interoperability across road, rail and maritime freight systems.
AI-enabled predictive supply chain resilience
In an increasingly volatile global economy, supply chain disruptions caused by geopolitical conflicts, extreme weather events, port congestion and fluctuating demand patterns are major challenges for fleet operators. Traditional logistics systems rely on historical data and real-time tracking to respond to disruptions after they occur, often resulting in costly delays and inefficiencies. AI is transforming supply chain management from a reactive model to a predictive and adaptive system, fundamentally redefining how fleets deal with uncertainty (Du Plessis et al., 2025[76]).
The emerging application of AI in predictive supply chain resilience is distinct from traditional AI-driven route optimisation, which focuses on finding the most efficient path at a given moment. Predictive resilience AI anticipates disruptions, recommending adjustments weeks or even months in advance. ML models continuously ingest real-time geopolitical intelligence, weather forecasts, economic trends and infrastructure conditions, allowing fleets to identify vulnerabilities before they affect operations (Aljohani, 2023[84]; Du Plessis et al., 2025[76]). For example, an AI system may detect early warning signals of a hurricane forming along a major freight route or an economic downturn affecting fuel prices. Rather than reacting only once disruptions occur, AI-powered systems support contingency planning. Identifying alternative routes or adjusting logistics schedules, for example, can minimise operational delays and financial impact, while respecting external constraints (Aljohani, 2023[84]).
Another key aspect of predictive resilience is AI-driven inventory and demand forecasting, which extends beyond individual fleet operations to optimise the entire supply chain network. Traditional forecasting relies on seasonal trends and historical demand patterns. Conversely, AI integrates real-time market data, warehouse capacity and transport availability to provide adaptive, demand-sensitive logistics planning. This ensures that fleets allocate the right vehicles to the right routes at the right time, reducing idle capacity and improving efficiency (Albayrak Ünal, Erkayman and Usanmaz, 2023[85]).
This emerging use-case differs from current AI-based logistics planning, which optimises within fixed parameters. Predictive supply chain resilience AI is dynamic as it continuously recalibrates based on new risk factors, shifting regulations and emerging disruptions. In so doing, it allows fleets to stay ahead of challenges rather than simply respond to them.
AI-driven sustainability analytics and green fleet optimisation
The role of AI in sustainability is expanding from basic fuel efficiency improvements to comprehensive, system-wide environmental management. With growing regulatory pressures on emissions reduction and the urgent need for carbon-neutral logistics, AI is being leveraged to actively manage fleet sustainability, integrate electric and hybrid vehicles, and provide data-driven emissions compliance strategies (Tanasuica and Roman, 2024[86]).
This emerging AI application is distinct from traditional fleet efficiency models, which primarily minimise fuel consumption and optimise routing for cost savings. Instead, AI-driven sustainability analytics focus on measuring, reporting and mitigating the overall environmental impact of transport operations (Chen et al., 2024[87]).
AI-powered emissions monitoring platforms collect real-time CO₂ data from vehicle exhaust sensors, telematics logs and GPS mileage tracking, providing instant sustainability reports. Unlike conventional annual emissions audits, AI allows fleets to continuously adjust operations to align with global carbon reduction targets. Some AI models even provide automated carbon credit calculations, helping fleets participate in carbon trading markets to offset their emissions (Chen et al., 2024[87]).
Another expanding application is AI-powered green fleet transition management. As logistics companies move towards electric and hybrid fleets, AI is used to optimise planning of EV fleet deployment, monitor battery health and allocate charging infrastructure. AI models analyse charging infrastructure availability, route topography and vehicle load demands, ensuring that EVs are deployed efficiently and charging downtime is minimised. For example, Nowos, a Dutch battery start-up, applies AI-driven diagnostics to monitor the health of lithium-ion batteries in EVs. Similarly, VivaDrive, a Poland-based AI company, uses digital twin modelling and fleet data analytics to help operators transition to low-emission fleets (Box 5.4).
As transport operators increasingly adopt electric vehicles (EVs), AI-based solutions play a growing role in addressing the challenges of fleet electrification and sustainability. Start-ups like VivaDrive and Nowos offer AI-driven tools to support battery lifecycle management, emissions monitoring and fleet optimisation.
VivaDrive (Poland)
VivaDrive, a Polish software start-up, offers an AI-powered platform to help organisations manage the transition to low-emission and EV fleets. The platform aggregates telematics, energy consumption and operational data to generate emissions reports and provide insights into overall fleet performance. Using AI-based simulations, it models different electrification scenarios to support decision making on vehicle selection, charging infrastructure and operational planning. The tool also offers route optimisation features and provides fleet managers with recommendations to improve energy efficiency and comply with emissions regulations.
Nowos (the Netherlands)
Nowos, a Dutch start-up founded in 2019, applies AI-based diagnostics to monitor the health of lithium-ion batteries in electric fleets. By analysing sensor data and usage patterns, the system identifies early signs of battery degradation and predicts maintenance needs. This enables fleet operators to make data-informed decisions about battery reuse, refurbishment or recycling, contributing to longer battery life spans and reducing replacement costs. In addition to its software, Nowos operates battery repair hubs across Europe that inspect and service batteries to prepare them for reuse. In April 2025, Nowos secured EUR 6 million to expand its European hub network, aiming to meet growing demand for battery maintenance and repair services.
Sources: Lawrence (2025[88]), “NOWOS raises €6M to expand lithium-ion battery repair and maintenance hubs”, https://tech.eu/2025/04/09/nowos-raises-eur6m-to-expand-lithium-ion-battery-repair-and-maintenance-hubs/; Nowos (2025[89]), “We bring new energy to batteries”, https://www.nowos.com/; VivaDrive (2025[90]), “Deploy and manage electric vehicles in your fleet”, https://vivadrive.io/.
Unlike current fuel-saving AI models, which optimise internal combustion engine vehicles, AI-driven sustainability analytics take a holistic approach to decarbonisation, integrating EV management, alternative fuels and regulatory compliance into fleet-wide decision making. As part of its broader digitalisation strategy, for example, the Port of Rotterdam in the Netherlands has deployed AI-based systems for smart container and stowage management. These systems leverage AI to predict optimal loading and unloading sequences based on factors such as vessel schedules, container weight and destination. This has contributed to a more efficient flow of goods through the port, helping reduce congestion and minimise environmental impacts from idling vessels (digitaldefynd, 2025[78]).
AI in multimodal freight co‑ordination
The co‑ordination of goods across multiple transport modes such as road, rail and maritime remains one of the most complex areas in freight logistics. AI has the potential to address long-standing co‑ordination challenges by supporting real-time decision making and adaptive planning across modal interfaces. While traditional freight operations often rely on siloed systems and fragmented infrastructure, AI tools could help improve synchronisation by integrating data from different logistics actors and transport operators (Dzemydienė, Burinskienė and Miliauskas, 2021[91]; Du Plessis et al., 2025[76]).
Examples of AI-enabled applications in this domain include real-time cargo tracking, estimated time of arrival prediction across modal legs, container allocation and disruption-sensitive scheduling. ML and reinforcement learning techniques can be used to forecast demand fluctuations and optimise cross-modal scheduling. To that end, they can account for variables such as port congestion, rail capacity or weather-related delays (Aljohani, 2023[84]; Chen et al., 2024[32]). AI-driven co‑ordination platforms may support logistics providers in anticipating bottlenecks and adjusting routings in advance, potentially increasing system resilience.
Moreover, AI has been linked to ongoing efforts to shift freight from road to rail or waterways, particularly where optimisation tools can make alternative routes more viable and competitive. Some studies suggest that use of AI for improved visibility and interoperability could contribute to broader decarbonisation and resilience goals (Durlik et al., 2024[75]). However, these efforts are ultimately contingent on physical infrastructure and connectivity in the built environment, which determine the feasibility of multimodal routing and transfers. Real-world implementation also depends on trust and co‑operation across actors, as well as on shared access to high-quality, standardised data. Persistent gaps in data interoperability, institutional alignment and business incentives continue to limit the large-scale uptake of AI tools for cross-modal freight planning (Dzemydienė, Burinskienė and Miliauskas, 2021[91]; Du Plessis et al., 2025[76]).
AI for automated fleet co‑ordination
As automated fleets – comprising trucks, yard vehicles or drones – become more technically feasible, the question of how to co‑ordinate such systems at scale has gained increasing attention. AI may play an important role in enabling co‑ordination of automated fleets. This is especially the case in complex logistics environments where multiple automated agents must interact with infrastructure, human operators and one another. While individual autonomy has progressed significantly, the effective orchestration of automated fleets remains a key challenge, and AI-based approaches are being explored as part of the solution (Shi, 2022[80]; Tanasuica and Roman, 2024[86]).
AI tools could support use-cases such as route allocation, task distribution, vehicle platooning and traffic-aware dispatching. In high-throughput settings like port terminals or distribution centres, multi-agent reinforcement learning systems may help co‑ordinate movements and avoid congestion. In parallel, predictive algorithms are being developed to help logistics operators assign tasks to the most suitable vehicle based on availability, charging needs or route conditions (Albayrak Ünal, Erkayman and Usanmaz, 2023[85]; Du Plessis et al., 2025[76]).
Nonetheless, the deployment of such systems remains uneven. Studies point to regulatory uncertainty (e.g. about applicable rules, or appropriate ways to demonstrate compliance); liability concerns (e.g. difficulties understanding how to determine and apportion responsibility in case of accident, harm or failure) and the lack of interoperable control platforms as major barriers to adoption. While promising pilots exist, particularly in closed environments or controlled logistics corridors, broader deployment of AI-based fleet co‑ordination is likely to depend on infrastructure readiness, standard-setting efforts and business model viability.
Insights from workshops and interview
Most commonly discussed use-cases
AI technologies increasingly support a range of operational functions in freight transport, including route optimisation, PdM, cargo safety and documentation management.
The ICS, one of the world’s main shipping organisations, assessed the quality of adoption of AI applications across its members (Table 5.4).
Source: Interview with ICS.
Among the most mature and widely adopted use-cases is AI-assisted route planning. This enables operators to reduce fuel consumption, reduce GHG emissions and improve operational efficiency. Systems such as Blue Visby (2025[92]) are being tested to facilitate just-in-time arrivals at ports and reduce unnecessary idling at anchor.
PdM and cargo monitoring are also gaining traction, although adoption remains concentrated among large operators. AI systems can analyse equipment usage and historical performance data to anticipate failures and support more efficient maintenance planning, particularly for safety-critical components. As illustrated by Maersk’s experience, PdM has helped reduce unplanned downtime and cut maintenance costs by up to 20%. However, smaller operators may lack the data scale or resources to develop such systems in-house, often relying instead on off-the-shelf software solutions. In cargo management, AI may help identify mis-declared or undeclared dangerous goods, which remains a persistent challenge on container vessels. These applications improve cargo segregation strategies and contribute to fire and explosion prevention on board.
At the port level, AI is being used to automate documentation processes (e.g. digital consignment notes), optimise crane and traffic light operations, and improve co‑ordination across complex multimodal logistics chains. Use-cases shared during the workshop illustrated how AI is being integrated into systems for traffic signal priority, speed advisory, fraud prevention and document handling, potentially increasing transparency, safety and reliability in logistics operations. While widespread digitalisation remains uneven across the sector, advanced ports such as Rotterdam, Hamburg and Valencia are using AI to co‑ordinate traffic flows, allocate docking slots and integrate rail and road freight logistics. Industry interviews also highlighted emerging applications in fraud detection, traffic light optimisation and crew work-rest planning systems that enhance compliance and safety management.
Despite these advances, adoption remains highly uneven across operators, reflecting differences in scale, resources and data availability. Smaller shipping companies, often family-run and operating only a few vessels, are less likely to deploy proprietary AI models, instead relying on external software providers.
Participants from the ERTICO workshop also showcased how AI for smart infrastructure management is advancing through tools such as Green Light Optimal Speed Advisory and AI-enhanced traffic light priority systems, as demonstrated by NeoGLS. These solutions provide real-time speed recommendations to reduce emissions from unnecessary stopping, while dynamically adjusting signal timing based on traffic composition (e.g. freight vehicles, buses, cyclists). Following successful pilots, they are now being deployed at scale across 15 testbeds in the European Union.
AI is also being applied to improve safety, compliance and emergency response in the transport of dangerous goods. NeoGLS presented use-cases in which AI enhances electronic freight transport information (eFTI) systems by automating document checks, issuing early risk alerts and facilitating cross-border data integration. These tools are central to France’s national eFTI infrastructure and enable more efficient communication between drivers, customs officials and emergency responders.
Use-cases discussed at the EC workshop on AI in Mobility and Transport also include intelligent logistics control platforms such as Fraunhofer’s “Omnistics”, which aggregate and analyse operational data across the supply chain. These platforms integrate data from internal systems, external partners and IoT sensors to support real-time decision making, quality control and fraud detection. Natural language AI assistants like LoOmni-Chat are being tested to further streamline user interactions, document handling and compliance management.
In parallel, Fraunhofer’s “Intelligent Digital Twin” framework demonstrates how AI can be used to simulate freight scenarios, predict disruptions and improve planning. By combining real-time data with simulation models, operators can test “what-if” logistics strategies before implementation. These range from emissions reduction planning to yard equipment optimisation and multimodal disruption forecasting.
Key barriers and challenges
Adoption across the freight transport sector remains uneven due to a mix of technical, economic and organisational barriers. Some larger operators are advancing AI integration through in-house development or by procuring AI systems and leveraging rich operational datasets. However, smaller operators often lack the scale, data availability or technical capacity to do so. The structural fragmentation of sectors such as maritime transport – where many companies own only a small number of vessels – can limit the volume and consistency of operational data needed to support AI training and deployment. In addition, operators are often reluctant to share data due to its perceived commercial value; the absence of secure sharing frameworks; and data protection regulations.
A second major barrier relates to low level of digitalisation and the limited interoperability of digital systems. According to workshop participants, more than 75% of transport documents are still processed manually, which poses a barrier to implementing AI tools that rely on structured data inputs. Addressing this gap may require foundational investment in digitalisation. Conversely, it may demand use of AI solutions such as intelligent document capture and automated data extraction to accelerate the shift from analogue to digital processes.
Another recurrent concern highlighted in the workshops was the lack of trust and alignment between stakeholders, particularly in the context of data exchange. Stakeholders reported institutional silos, unclear ownership rules and regulatory inconsistencies across Member States as major inhibitors to building collaborative, AI-ready logistics ecosystems.
Outlook/other AI-based solutions
Rather than displacing human judgement, AI was framed as a tool to augment operational awareness and support data-informed decisions, particularly in complex, multimodal logistics environments. Participants noted that transformative change would require foundational shifts in both technological and organisational readiness.
Participants noted a growing focus on AI-enabled co‑ordination across multimodal freight chains, the integration of AI into port operations and the potential for AI-driven sustainability tracking. However, several stakeholders emphasised that scaling up these innovations will depend on addressing foundational gaps in digitalisation, data standardisation and interoperability, especially for small and mid-sized operators. Trust-based data ecosystems and more collaborative approaches to data sharing were seen as critical enablers of progress.
Key recommendations
Industry stakeholders and workshop participants put forward several recommendations to support the wider and more effective deployment of AI in freight transport:
Data
Foster interoperability for multimodal freight co‑ordination: promote secure, purpose-specific data exchange by enhancing interoperability between platforms and systems, while addressing commercial sensitivities, data privacy and trust concerns to enable effective AI‑supported co‑ordination across logistics operators.
Support investment in data standardisation and interoperability: facilitate seamless data integration across diverse systems and transport modes by promoting common standards and shared data governance frameworks.
Accelerate digital documentation adoption as a foundation for AI: promote the use of electronic consignment notes and automated document processing tools, especially for SMEs. Prioritise funding to help operators transition from paper-based systems and ensure structured data flows.
Infrastructure and connectivity
Provide funding for digital and AI-ready logistics infrastructure upgrades: create dedicated funding programmes to support the digitalisation and retrofitting of ports, terminals and logistics hubs, laying the groundwork for AI-powered automation, safety systems and data connectivity.
AI solutions, software and interoperability
Promote modular, interoperable AI solutions: foster development of modular, flexible AI applications that SMEs and smaller operators can adopt without requiring costly customisation or vendor lock-in.
Skills, talent and collaboration
Invest in workforce upskilling in AI and digital transformation: develop and fund training programmes that build AI literacy, data science skills and digital capabilities among logistics workers and managers, enabling effective adoption and use of AI technologies.
Foster public-private-academic partnerships for AI development: support collaborative initiatives that co-develop AI solutions tailored to sector-specific challenges in freight and logistics, ensuring tools address real-world operational needs.
AI-related skills
To ensure that AI can be deployed effectively and responsibly in the mobility sector, a broad range of technical, operational and organisational skills are required. Industry stakeholders underlined the need for multidisciplinary expertise that bridges AI development with deep knowledge of transport systems, regulatory frameworks and human-centred design. Drawing on insights from workshops and interviews, the following areas of expertise appear particularly relevant for supporting the development and deployment of AI in mobility.
Expertise in AI model development and data science: a foundational skill is advanced knowledge of AI development techniques, including ML, deep learning and reinforcement learning. Specialists are needed to design, train and deploy models for key mobility applications such as object detection, trajectory planning, traffic optimisation and PdM. Robust data science capabilities, including expertise in data wrangling, time-series analysis and multimodal data processing, are essential for building reliable, real-time AI systems.
Knowledge of sensor integration and perception systems: automated driving, smart infrastructure and AI-enhanced fleet management all rely on complex sensor ecosystems. Skills in developing sensor fusion algorithms and in integrating LiDAR, radar, cameras and other IoT sensors are crucial to ensure accurate perception, localisation and situational awareness across diverse mobility environments.
Competence in AI ethics and regulatory compliance: stakeholders emphasise the growing need for expertise in ethical AI principles, impact assessments and bias mitigation strategies. Professionals must be equipped to ensure that AI systems in mobility uphold fairness, transparency and accountability and also comply with frameworks such as the AI Act, the safety standards of the United Nations Economic Commission for Europe and sector-specific guidelines. This skillset will be increasingly relevant as regulatory requirements for trustworthy AI evolve.
Expertise in cybersecurity, data privacy and data governance: connected vehicles, smart traffic systems and AI-driven logistics require robust data governance frameworks. Skills in securing AI systems through cybersecurity measures, data encryption and anomaly detection are essential. Equally important is knowledge of GDPR compliance, ethical data use and emerging regulations governing AI in transport.
Familiarity with transport systems and systems integration: AI solutions in mobility must be closely aligned with the operational realities of transport networks. This requires strong sectoral expertise – whether in public transport, freight logistics or traffic systems – along with deep understanding of multimodal co‑ordination, fleet management and infrastructure. Equally essential are skills in systems integration. AI in mobility typically interacts with infrastructure, vehicle systems and digital platforms, making expertise in systems architecture, API development and interoperability standards crucial to ensure seamless integration, cross-platform communication and scalability of AI tools. In parallel, public authorities must also build institutional capacity and AI literacy among their staff, enabling them to procure, oversee and deploy AI technologies in line with public service objectives effectively.
Capabilities in organisational transformation and change management: beyond technical deployment, the success of AI in mobility depends on an organisation’s ability to adapt. Skills in change management, staff training and process redesign are vital to help transport operators reconfigure workflows, build internal buy-in and address resistance to new technologies. These capabilities ensure that AI tools deliver lasting value, particularly in public transport and freight sectors, where legacy systems and institutional inertia can slow innovation.
Competence in testing, validation and certification: ensuring the safety, reliability and accountability of AI systems requires specialised skills in simulation, validation and certification. Engineers and data scientists must be trained in designing safety cases, conducting edge-case testing and navigating the technical and procedural requirements for certifying AI systems, especially in safety-critical domains such as AVs and transport infrastructure.
Expertise in human-machine interaction and user-centred design: building trust in AI-powered mobility systems depends on intuitive, transparent and adaptive interfaces. Skills in human-machine interaction, user experience (UX) design and behavioural science are critical to ensure that AI systems can communicate effectively with drivers, passengers, operators and maintenance teams, reducing cognitive load and fostering acceptance.
Key recommendations to enhance AI uptake in mobility in the European Union
Data availability and access
Foster secure, interoperable data sharing: promote EU-level data pooling and sharing mechanisms, including by leveraging the common European mobility data spaces. Address issues of data ownership, privacy and competition by incentivising secure, privacy-preserving data-sharing protocols to encourage participation across sectors.
Invest in data standardisation and integration: fund the development and promote the alignment of standards and common technical frameworks to support the seamless exchange of data across public transport, AVs and multimodal logistics.
Build collaborative data platforms: support development of shared data management and integration platforms, providing access to clean and standardised data and facilitating co‑operation between public and private actors.
Infrastructure and connectivity
Upgrade infrastructure: provide targeted funding programmes and encourage private investments to retrofit roads, terminals and facilities to support safe, efficient, AI-enabled use-cases (e.g. AV‑signage friendly signage, connectivity at logistics hubs).
Expand digital and physical AI-ready digital and physical infrastructure: increase investment in V2X systems, 5G networks, IoT-enabled infrastructure, and AI-compatible equipment across urban areas and logistics corridors, and road networks to enable connected and automated mobility applications.
Bridge the urban-rural divide: offer dedicated funding, grants or vouchers to equip smaller cities, ports and terminals with the infrastructure needed for AI adoption.
AI solutions, software and interoperability
Promote open-source, modular and interoperable AI ecosystems: encourage development of open AI algorithms, model weights and interoperable software to reduce vendor lock-in, support SME adoption and stimulate innovation across the transport sector.
Enable experimentation through pilots and virtual testing: fund pilot programmes and support digital twin technologies to allow public and private actors to test AI solutions in controlled environments, evaluate their impact and reduce investment risk before large-scale deployment.
Foster development of sector-specific LLMs: support the development and training of domain-specific LLMs tailored to the mobility and transport sector, leveraging proprietary European datasets to build competitive and trustworthy AI solutions.
Regulatory frameworks and testing environments
Support implementation guidance and regulatory clarity: ensure that the AI Act and related regulations are accompanied by clear, practical guidance for transport-sector stakeholders, helping organisations understand compliance requirements, enforcement expectations and permissible experimentation pathways.
Establish regulatory sandboxes and testbeds: support the creation of cross-border regulatory testbeds and national sandboxes where cities, regions and logistics operators can experiment with AI solutions, waiving compliance burdens in controlled settings and facilitating innovation while safeguarding public interest.
Skills, talent and collaboration
Invest in AI workforce development and upskilling: support EU-wide initiatives and targeted programmes to train technical and non-technical staff in AI, data science and digital transformation, addressing skills gaps across both public and private actors in transport and mobility sectors.
Strengthen public-private-academic innovation partnerships: fund collaborative research and innovation hubs that bring together transport authorities, industry, start-ups and universities to co‑develop AI applications addressing operational challenges in the mobility sector.
Promote peer learning and public trust: establish knowledge-sharing networks, exchanges of best practices and awareness campaigns to foster public trust in AI applications, address societal concerns (around privacy, safety and employment) and ensure inclusive adoption of AI-driven mobility solutions across Member States.
References
[39] 5GMOBIX (2025), “5GMOBIX – Breaking boundaries with 5G”, https://www.5g-mobix.com/ (accessed on 13 February 2025).
[83] Abduljabbar, R. et al. (2019), “Applications of artificial intelligence in transport: An overview”, Sustainability, Vol. 11/1, p. 189, https://doi.org/10.3390/su11010189.
[57] Abirami, S. et al. (2024), “A systematic survey on big data and artificial intelligence algorithms for intelligent transportation system”, Case Studies On Transport Policy, Vol. 17, p. 101247, https://doi.org/10.1016/j.cstp.2024.101247.
[51] Adanyin, A. and J. Odede (2024), “AI-driven fare evasion detection in public transportation: A multi-technology approach integrating behavioural AI, IoT, and privacy-preserving systems”, Preprints.org, https://doi.org/10.20944/preprints202412.0127.v1.
[61] AIAMO (2025), “Artificial intelligence and mobility”, https://www.aiamo.de/en (accessed on 13 February 2025).
[85] Albayrak Ünal, Ö., B. Erkayman and B. Usanmaz (2023), “Applications of artificial intelligence in inventory management: A systematic review of the literature”, Archive of Computational Methods in Engineering, Vol. 30, pp. 2605-2625, https://doi.org/10.1007/s11831-022-09879-5.
[84] Aljohani, A. (2023), “Predictive analytics and machine learning for real-time supply chain risk mitigation and agility”, Sustainability, Vol. 15/20, p. 15088, https://doi.org/10.3390/su152015088.
[73] Alqasi, A., Y. Alkelanie and A. Alnagrat (2024), “Intelligent infrastructure for urban transportation: The role of artificial intelligence in predictive maintenance”, Brilliance: Research of Artificial Intelligence, Vol. 4, pp. 625-637, https://doi.org/10.47709/brilliance.v4i2.4889.
[24] ASIMOB (2025), “Autonomous road inspector”, https://asimob.es/en/ (accessed on 15 February 2025).
[18] Atakishiyev, S. et al. (2024), “Explainable artificial intelligence for autonomous driving: A comprehensive overview and field guide for future research directions”, IEEE Access, Vol. 12, pp. 101603-101625, https://doi.org/10.1109/ACCESS.2024.3431437.
[64] AWAAIT (2025), “Subway fare evasion detection”, https://www.awaait.com/products (accessed on 17 February 2025).
[48] Axelera AI (2025), “Home page”, https://axelera.ai/ (accessed on 20 February 2025).
[63] Bianchi, G. et al. (2025), “Systematic review railway infrastructure monitoring: From classic techniques to predictive maintenance”, Advances in Mechanical Engineering, Vol. 17/1, p. 16878132241285631, https://doi.org/10.1177/16878132241285631.
[92] Blue Visby (2025), “Blue Visby solution”, https://bluevisby.com/ (accessed on 5 May 2025).
[33] Bram-Larbi, K. et al. (2020), “Improving emergency vehicles’ response times with the use of augmented reality and artificial intelligence”, HCI International 2020 – Late Breaking Papers: Digital Human Modeling and Ergonomics, Mobility and Intelligent Environments. HCII 2020. Lecture Notes in Computer Science, pp. 24-39, https://doi.org/10.1007/978-3-030-59987-4_3.
[81] CGA CGM Group (2025), “The CMA CGM Group adopts custom-designed AI solutions from Mistral AI to support its shipping, logistics, and media activities”, https://www.cmacgm-group.com/en/news-media/cma-cgm-group-adopts-custom-designed-ai-solutions-mistral-ai (accessed on 20 February 2025).
[32] Chen, L. et al. (2024), “End-to-end autonomous driving: Challenges and frontiers”, arXiv, Vol. 2306.16927, https://doi.org/10.48550/arXiv.2306.16927.
[87] Chen, W. et al. (2024), “Artificial intelligence in logistics optimization with sustainable criteria: A review”, Sustainability, Vol. 16/21, p. 9145, https://doi.org/10.3390/su16219145.
[31] Chib, P. and P. Singh (2023), “Recent advancements in end-to-end autonomous driving using deep learning: A survey”, IEEE Transactions on Intelligent Vehicles, Vol. 9/1, pp. 103-118, https://doi.org/10.1109/TIV.2023.3318070.
[65] Chu, K. et al. (2024), “A survey of artificial intelligence-related cybersecurity risks and countermeasures in mobility-as-a-service”, IEEE Intelligent Transportation Systems Magazine, Vol. 16/6, pp. 37-55, https://doi.org/10.1109/MITS.2024.3427655.
[54] Cohen, A. (2024), The Role of Artificial Intelligence in Transportation, San José State University and Mineta Transportation Institute, https://transweb.sjsu.edu/sites/default/files/2430-Cohen-Artificial-Intelligence-Transportation-Use-Cases.pdf.
[23] Cui, C. et al. (2024), “A survey on multimodal large language models for autonomous driving”, Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 958-979, https://openaccess.thecvf.com/content/WACV2024W/LLVM-AD/html/Cui_A_Survey_on_Multimodal_Large_Language_Models_for_Autonomous_Driving_WACVW_2024_paper.html.
[66] deployEMDS (2025), “Barcelona”, https://deployemds.eu/deployment/barcelona/ (accessed on 17 June 2025).
[25] Detekt (2025), “AI for road asset management with mobile mapping data”, https://www.detekt.it/?utm_source=website&utm_medium=button&utm_campaign=enliteAI (accessed on 27 February 2025).
[78] digitaldefynd (2025), “How can AI be used in the shipping industry [10 case studies]”, https://digitaldefynd.com/IQ/ai-use-in-the-shipping-industry-case-studies/ (accessed on 24 March 2025).
[50] Ding, Z. et al. (2023), “An artificial intelligence-based method for crack detection in engineering facilities around subways”, Applied Sciences, Vol. 13/19, p. 11002, https://doi.org/10.3390/app131911002.
[76] Du Plessis, M. et al. (2025), “Shaping the future of freight logistics: Use cases of artificial intelligence”, Sustainability, Vol. 17/4, p. 1355, https://doi.org/10.3390/su17041355.
[75] Durlik, I. et al. (2024), “Artificial intelligence in maritime transportation: A comprehensive review of safety and risk management applications”, Applied Sciences, Vol. 14/18, p. 8420, https://doi.org/10.3390/app14188420.
[91] Dzemydienė, D., A. Burinskienė and A. Miliauskas (2021), “Integration of multi-criteria decision support with infrastructure of smart services for sustainable multi-modal transportation of freights”, Sustainability, Vol. 13/9, p. 4675, https://doi.org/10.3390/su13094675.
[4] EEA (2025), “Transport and mobility”, European Environment Agency, https://www.eea.europa.eu/en/topics/in-depth/transport-and-mobility?activeTab=07e50b68-8bf2-4641-ba6b-eda1afd544be (accessed on 24 March 2025).
[26] enlite (2025), “Bringing artificial intelligence to your organization”, https://www.enlite.ai/ (accessed on 12 March 2025).
[40] ERTICO (2025), “ERTICO focus on event on data and AI for safe and resilient ITS”, European Road Transport Telematics Implementation Coordination Organisation, https://erticonetwork.com/event/ertico-focus-on-event-on-data-and-ai-for-safe-and-resilient-its/ (accessed on 5 May 2025).
[46] European Commission (2025), “Creating a common European mobility data space”, https://transport.ec.europa.eu/transport-themes/smart-mobility/creating-common-european-mobility-data-space_en (accessed on 17 June 2025).
[45] European Commission (2025), “deployEMDS”, https://deployemds.eu/ (accessed on 14 August 2025).
[7] European Commission (2025), Industrial Action Plan for the European Automotive Sector, COM(2025) 95 final, European Commission, Brussels, https://transport.ec.europa.eu/document/download/89b3143e-09b6-4ae6-a826-932b90ed0816_en?filename=Communication+-+Action+Plan.pdf.
[41] European Commission (2025), “Workshop – AI in automotive: Barriers, challenges and opportunities”, https://digital-strategy.ec.europa.eu/en/events/ai-automotive-applications-opportunities-and-barriers (accessed on 5 May 2025).
[22] European Commission (2024), Status of Progress on Connected, Cooperative and Automated Mobility in Europe, European Commission, Brussels, https://research-and-innovation.ec.europa.eu/document/download/1720a5ef-01bf-498e-85f5-c61bb3a7bc31_en?filename=swd_2024_92.pdf.
[42] European Commission (2023), “Creation of a common European mobility data space”, COM(2023) 751 final, European Commission, Brussels, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM%3A2023%3A751%3AFIN.
[43] European Commission (2023), “The completion of PrepDSpace4Mobility”, https://digital-strategy.ec.europa.eu/en/news/completion-prepdspace4mobility (accessed on 14 August 2025).
[6] European Commission (2020), Sustainable & Smart Mobility Strategy, European Commission, Brussels, https://transport.ec.europa.eu/document/download/be22d311-4a07-4c29-8b72-d6d255846069_en.
[5] European Commission (2019), “The European Green Deal”, COM(2019) 640 final, European Commission, Brussels, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM%3A2019%3A640%3AFIN.
[13] European Parliament & Council (2023), “Directive (EU) 2023/2661”, https://eur-lex.europa.eu/eli/dir/2023/2661/oj/eng (accessed on 20 June 2025).
[9] Eurostat (2025), “Artificial intelligence by NACE Rev. 2 activity”, in Science, technology, digital society, Eurostat, https://doi.org/10.2908/ISOC_EB_AIN2.
[1] Eurostat (2025), “Key figures on European transport – 2024 edition”, https://ec.europa.eu/eurostat/web/products-key-figures/w/ks-01-24-021 (accessed on 13 March 2025).
[3] Eurostat (2025), “Statistics explained: Use of artificial intelligence in enterprises”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises (accessed on 13 March 2025).
[2] Eurostat (2024), “Freight transport statistics – modal split”, https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Freight_transport_statistics_-_modal_split (accessed on 13 March 2025).
[19] Garikapati, D. and S. Shetiya (2024), “Autonomous vehicles: Evolution of artificial intelligence and the current industry landscape”, Big Data and Cognitive Computing, Vol. 8/4, p. 42, https://doi.org/10.3390/bdcc8040042.
[62] Güven, Ö. and H. Sahin (2022), “Predictive maintenance based on machine learning in public transportation vehicles”, Mühendislik Bilimleri ve Araştırmaları Dergisi, Vol. 4/1, pp. 89-98, https://doi.org/10.46387/bjesr.1093519.
[47] HammerHAI (2025), “HammerHAI – An AI factory for manufacturing, engineering and research”, https://www.hammerhai.eu/ (accessed on 5 May 2025).
[27] Hu, A. et al. (2022), “Model-based imitation learning for urban driving”, Advances in Neural Information Processing Systems, Vol. 35, pp. 20703-20716, https://proceedings.neurips.cc/paper_files/paper/2022/hash/827cb489449ea216e4a257c47e407d18-Abstract-Conference.html.
[29] ITF (2024), “Lost in transmission: Communicating for safe automated vehicle interactions in cities”, International Transport Forum, https://www.itf-oecd.org/communicating-safe-automated-vehicle-cities (accessed on 17 June 2025).
[52] Jevinger, A. et al. (2024), “Artificial intelligence for improving public transport: A mapping study”, Public Transport, Vol. 16/1, pp. 99-158, https://doi.org/10.1007/s12469-023-00334-7.
[16] Khayyam, H. et al. (2019), “Artificial intelligence and Internet of Things for autonomous vehicles”, in Nonlinear Approaches in Engineering Applications, Springer, Cham, https://doi.org/10.1007/978-3-030-18963-1_2.
[88] Lawrence, C. (2025), “NOWOS raises €6M to expand lithium-ion battery repair and maintenance hubs”, tech.eu, https://tech.eu/2025/04/09/nowos-raises-eur6m-to-expand-lithium-ion-battery-repair-and-maintenance-hubs/ (accessed on 24 March 2025).
[55] Liyanage, S. et al. (2022), “AI-based neural network models for bus passenger demand forecasting using smart card data”, Journal of Urban Management, Vol. 11/13, pp. 365-380, https://doi.org/10.1016/j.jum.2022.05.002.
[14] Ma, Y. et al. (2020), “Artificial intelligence applications in the development of autonomous vehicles: A survey”, IEEE/CAA J. Autom. Sinica, Vol. 7/2, pp. 315-329, https://doi.org/10.1109/JAS.2020.1003021.
[71] metaCCAZe (2025), “Welcome to metaCCAZe!”, https://www.metaccaze-project.eu/ (accessed on 14 August 2025).
[69] Mirindi, D. (2024), “A review of the advances in artificial intelligence in transportation system development”, Journal of Civil, Construction and Environmental Engineering, Vol. 9/3, pp. 72-83, https://doi.org/10.11648/j.jccee.20240903.13.
[35] Mishra, S. and S. Das (2019), “A review on vision based control of autonomous vehicles using artificial intelligence techniques”, 2019 International Conference on Information Technology (ICIT), pp. 500-504, https://doi.org/10.1109/ICIT48102.2019.00094.
[72] MOBILITIES FOR EU (2025), “MOBILITIES FOR EU – for a better future”, https://mobilities-for.eu/ (accessed on 14 August 2025).
[17] Muhammad, K. et al. (2021), “Deep learning for safe autonomous driving: Current challenges and future directions”, IEEE Transactions on Intelligent Transportation Systems, Vol. 22/7, pp. 4316-4336, https://doi.org/10.1109/TITS.2020.3032227.
[36] Namburi, V. et al. (2024), “Integrating AI and cybersecurity: Advancing autonomous vehicle security and response mechanisms”, Conference: 2024 International Conference on Emerging Trends in Networks and Computer Communications (ETNCC), Windhoek, Namibia, https://doi.org/10.1109/ETNCC63262.2024.10767464.
[28] Nascimento, A. et al. (2019), “A systematic literature review about the impact of artificial intelligence on autonomous vehicle safety”, IEEE Transactions on Intelligent Transportation Systems, Vol. 21/12, pp. 4928-4946, https://doi.org/10.1109/TITS.2019.2949915.
[49] Nikitas, A. et al. (2020), “Artificial intelligence, transport and the smart city: Definitions and dimensions of a new mobility era”, Sustainability, Vol. 12/7, p. 2789, https://doi.org/10.3390/su12072789.
[15] Noviati, N. et al. (2024), “Artificial intelligence in autonomous vehicles: Current innovations and future trends”, International Journal of Cyber and IT Service Management, Vol. 4/3, pp. 97-104, https://doi.org/10.34306/ijcitsm.v4i2.161.
[89] Nowos (2025), “We bring new energy to batteries”, https://www.nowos.com/ (accessed on 25 March 2025).
[8] OECD (2022), “OECD Framework for the Classification of AI systems”, OECD Digital Economy Papers, No. 323, OECD Publishing, Paris, https://doi.org/10.1787/cb6d9eca-en.
[10] OECD.AI (2025), AI talent concentration by country and industry, data from LinkedIn Economic Graph, last updated 2025-04-07, (database), https://oecd.ai/en/data?selectedArea=ai-jobs-and-skills&selectedVisualization=ai-talent-concentration-by-country-and-industry&visualizationFiltersHash=eyJmaWx0ZXJzIjp7IkluZHVzdHJ5IjoiTWFudWZhY3R1cmluZyJ9fQ%3D%3D (accessed on 10 April 2025).
[12] OECD.AI (2025), “Total VC investments in AI by country and industry, data from Preqin, last updated 2025-02-18”, OECD.AI Policy Observatory, (database), https://oecd.ai/en/data?selectedArea=investments-in-ai-and-data&selectedVisualization=total-vc-investments-in-ai-by-country-and-industry (accessed on 6 May 2025).
[60] Official Journal of the European Union (2010), DIrective 2010/40/EU of the European Parliament and of the Council of 7 July 2010 on the framework for the deployment of intelligent transport systems in the field of road transport and for interfaces with other modes of transport, L 207/1, Official Journal of the European Union, https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32010L0040.
[79] Olugbade, S. et al. (2022), “A review of artificial intelligence and machine learning for incident detectors in road transport systems”, Mathematical and Computational Applications, Vol. 27/5, p. 77, https://doi.org/10.3390/mca27050077.
[37] Onur, F. et al. (2024), “Machine learning-based identification of cybersecurity threats affecting autonomous vehicle systems”, Computers & Industrial Engineering, Vol. 190, p. 110088, https://doi.org/10.1016/j.cie.2024.110088.
[56] Paiva, S. et al. (2021), “Enabling technologies for urban smart mobility: Recent trends, opportunities and challenges”, Sensors, Vol. 21/6, p. 2143, https://doi.org/10.3390/s21062143.
[20] Parekh, D. et al. (2022), “A review on autonomous vehicles: Progress, methods and challenges”, Electronics, Vol. 11/14, p. 2162, https://doi.org/10.3390/electronics11142162.
[58] Patil, D. et al. (2024), “Artificial intelligence and generative AI, such as ChatGPT, in transportation: Applications, technologies, challenges, and ethical considerations”, Trustworthy Artificial Intelligence in Industry and Society, pp. 185-232, https://doi.org/10.70593/978-81-981367-4-9_6.
[38] PoDIUM (2025), “PoDIUM – Accelerating the implementation of connected, cooperative and automated mobility technology”, https://podium-project.eu/ (accessed on 25 March 2025).
[59] PTV Group (2024), “Hamburg’s LSBG and PTV Group forge strategic partnership for sustainable, integrated, and technology-driven mobility”, https://www.ptvgroup.com/en/resources/news/company/hamburgs-lsbg-and-ptv-group-forge-strategic-partnership-sustainable (accessed on 25 March 2025).
[70] Rataj, M. et al. (2025), New and Emerging Transport Technologies and Trends in European Research and Innovation Projects 2024, Joint Research Centre, European Commission, Brussels, https://publications.jrc.ec.europa.eu/repository/handle/JRC140839.
[21] Reda, M. et al. (2024), “Path planning algorithms in the autonomous driving system: A comprehensive review”, Robotics and Autonomous Systems, Vol. 174, p. 104630, https://doi.org/10.1016/j.robot.2024.104630.
[34] Rigas, E., A. Billis and P. Bamidis (2022), “Can artificial intelligence enable the transition to electric ambulances?”, Challenges of Trustable AI and Added-Value on Health, Vol. 294, pp. 73-77, https://ebooks.iospress.nl/pdf/doi/10.3233/SHTI220399.
[44] Scholliers, J. et al. (2023), Study in support of the creation of the common European mobility data space (EMDS), Final report, European Commission, Publications Office of the European Union, https://doi.org/10.2832/5074808.
[80] Shi, J. (2022), “Research on optimization of cross‐border e‐commerce logistics distribution network in the context of artificial intelligence”, Mobile Information Systems, Vol. 1, p. 3022280, https://doi.org/10.1155/2022/3022280.
[86] Tanasuica, Z. and M. Roman (2024), “Machine learning for concrete sustainability improvement: Smart fleet management”, Eastern European Journal for Regional Studies (EEJRS), Vol. 10/1, pp. 79-97, https://www.ceeol.com/search/article-detail?id=1252477.
[77] Tang, R. et al. (2022), “A literature review of artificial intelligence applications in railway systems”, Transportation Research Part C: Emerging Technologies, Vol. 140, p. 103679, https://doi.org/10.1016/j.trc.2022.103679.
[67] Tarkiainen, M. et al. (2021), “AI-based vehicle systems for mobility-as-a-service application”, in Artificial Intelligence for Digitising Industry – Applications, River Publishers, https://www.taylorfrancis.com/.
[68] Turno, F. and I. Yatskiv (2023), “Mobility-as-a-service: literature and tools review with a focus on personalization”, Transport, Vol. 38/4, pp. 243-262, https://doi.org/10.3846/transport.2023.20997.
[53] UITP (2025), Artificial Intelligence in Public Transport, International Association of Public Transport, Brussels, https://cms.uitp.org/wp/wp-content/uploads/2025/03/20250325_Artificial-Intelligence-in-Public-Transport.pdf.
[90] VivaDrive (2025), “Deploy and manage electric vehicles in your fleet”, https://vivadrive.io/ (accessed on 14 March 2025).
[11] WEF (2025), Autonomous Vehicles: Timeline and Roadmap Ahead, World Economic Forum, Cologny, https://reports.weforum.org/docs/WEF_Autonomous_Vehicles_2025.pdf (accessed on 14 August 2025).
[74] Woschank, M., E. Rauch and H. Zsifkovits (2020), “A review of further directions for artificial intelligence, machine learning, and deep learning in smart logistics”, Sustainability, Vol. 12/9, p. 3760, https://doi.org/10.3390/su12093760.
[82] Yang, G. et al. (2024), “Comprehensive assessment of artificial intelligence tools for driver monitoring and analyzing safety critical events in vehicles”, Sensors, Vol. 24/8, p. 2478, https://doi.org/10.3390/s24082478.
[30] Zhou, X. et al. (2024), “Vision language models in autonomous driving: A survey and outlook”, IEEE Transactions on Intelligent Vehicles, https://doi.org/10.1109/TIV.2024.3402136.
← 1. Additional relevant EU AI projects include AITHENA, AI4CCAM, AUGMENTED CCAM, AWARE2ALL, and CONNECT.