Generative AI for Smarter CCAM: Enhancing Perception, Decision-Making, and Validation

The program supports the availability and integration of advanced, trustworthy and energy-efficient perception systems using Generative AI; enhanced safety for Vulnerable Road Users (VRUs) through improved perception and behavioural understanding; enhanced robustness of CCAM systems through GenAI-generated training, testing and validation scenarios; improved understanding of the relevance and limitations of Generative AI in CCAM applications; development of tools and harmonised approaches for mobility technology development, training and validation; robust environment perception and decision-making solutions; scenario generation for testing and validation; integration of GenAI technologies into existing development and validation approaches; guidelines addressing the benefits, limitations, transparency, accountability, gender bias and fairness of GenAI applications; collaboration with the European Software-defined Vehicle initiative; and coordination with the European Connected and Autonomous Vehicle Alliance (ECAVA).

The growing deployment of Level 3 and Level 4 automated vehicle services presents significant challenges in perception and decision-making, particularly in complex urban environments where operating conditions can change rapidly and new situations frequently arise. These challenges require low-latency solutions capable of improving responsiveness, situational awareness and safety in real-time conditions.

The initiative recognizes the need to address limitations related to latency, bandwidth consumption and energy usage associated with on-board computing systems. At the same time, enhanced security, privacy and reliability are required to support scene understanding and the prediction of near-future developments in traffic environments. These capabilities are particularly important when interacting with Vulnerable Road Users and ensuring safe deployment of CCAM-enabled mobility solutions.

Recent advances in Generative AI technologies have demonstrated significant potential for supporting CCAM applications. Earlier exploratory work has highlighted the possibility of using GenAI to generate edge-case scenarios for the development, training, testing and validation of automated mobility systems. Building upon these developments, the current initiative seeks to further advance and validate GenAI applications specifically tailored to the CCAM domain.

Projects are expected to develop tools and approaches that strengthen environment perception and decision-making capabilities across edge computing systems, on-board vehicle platforms, infrastructure systems and back-office environments. Particular attention is placed on accelerating reasoning processes, improving efficiency, enhancing cybersecurity and increasing the reliability of automated mobility applications. Initial applications may include path planning, support for Vulnerable Road User perception, behaviour prediction and intention recognition, as well as data-sharing approaches that enable greater reaction times in near-accident situations.

The initiative also promotes the use of advanced Generative AI technologies, including Large Language Models, Vision Language Models and Vision Language Action systems. These technologies can contribute to improved contextual reasoning, pattern recognition, sensory input interpretation and enriched environmental understanding, thereby supporting more effective and adaptable decision-making processes.

Another key area of activity involves the generation of realistic interaction scenarios between CCAM-enabled vehicles and other road users. These scenarios are expected to support testing and validation activities by extending existing datasets and introducing meaningful variations that reflect differences in infrastructure conditions, traffic environments and user behaviours.

Projects should further investigate the integration of Generative AI technologies into existing development, training and validation methodologies. This includes assessing both the benefits and limitations of these technologies, developing guidelines for their responsible use, and addressing issues related to transparency, accountability, fairness and bias. The resulting tools and approaches may support a wide range of CCAM technologies and systemic applications, including traffic management and remote control systems.

The initiative encourages collaboration with the European Software-defined Vehicle initiative through the adoption of existing interfaces and building blocks, while also promoting the development of new components that may contribute to future framework enhancements. Coordination with the European Connected and Autonomous Vehicle Alliance is also expected to ensure alignment with broader European automotive and mobility strategies.

The broader impacts of this initiative include improved mobility for people and goods under diverse operating conditions, greater interoperability and integration of CCAM solutions within transport ecosystems, reduced environmental impacts through sustainable mobility solutions, enhanced competitiveness through advanced technologies, stronger resilience of passenger and freight transport networks, improved public transport resilience through AI applications, and contributions toward reducing road fatalities and improving safety across transport systems.

The total funding available for this topic is €13,000,000, with an indicative funding level of around €6,500,000 per project.

Any legal entity, regardless of its place of establishment, including entities from non-associated third countries and international organisations, may participate provided that the conditions established under the Horizon Europe Regulation and the specific topic requirements are met. Applicants must register in the Participant Register, obtain a Participant Identification Code (PIC), and complete the required validation procedures before signing a grant agreement.

For more information, visit EC.