AI4CCAM: Advancing Trustworthy and Explainable AI for Safer Urban Mobility

In the evolving landscape of connected and automated mobility, artificial intelligence (AI) plays a decisive role in shaping the future of transportation. Yet, as vehicles become smarter, questions of trust, transparency, and ethics come to the forefront. The AI4CCAM project addresses these challenges head-on, developing trustworthy and explainable AI models designed to make roads safer for everyone, especially Vulnerable Road Users (VRUs) such as pedestrians and cyclists.

The Challenge of Trustworthy AI

AI4CCAM tackles one of the most pressing issues in automated mobility: how can AI-driven systems predict human behaviour in unpredictable urban environments, and do so safely, fairly, and transparently?

The project’s mission extends beyond technical performance. It incorporates ethics, safety, and user acceptance into every aspect of AI development, while working to mitigate bias, ensure cybersecurity, and build public trust. At its core, AI4CCAM is also creating an open and interoperable testing environment, supporting the broader effort to operationalise trustworthy AI for Connected, Cooperative, and Automated Mobility (CCAM).

A Dual Approach to Trustworthiness

AI4CCAM examines trust from two complementary angles to achieve these goals. The first focuses on how automated vehicles perceive and interpret their surroundings, including interactions with other road users. The second explores how VRUs react to approaching automated vehicles, using virtual reality (VR) scenarios to capture real-world human behaviour.

This comprehensive approach ensures that both sides of the interaction, human and machine, are understood, modelled, and improved upon.

From Simulation to Real-World Insight

At the heart of AI4CCAM lies a robust digital framework and methodology, designed to support a complete AI development pipeline. This includes:

• Scenario prescriptions based on established automotive standards and trustworthiness criteria.
• Test protocols that extend beyond traditional ADAS benchmarks to include AI-specific pass/fail performance criteria.
• AI model development using open datasets and advanced training methods.
• Simulation in a Digital Twin environment (using the CARLA simulator) to test and validate models under realistic urban conditions.

In parallel, the project has built an online platform, the Participatory Space, to engage stakeholders, enabling co-creation and dialogue on ethical, social, and technical dimensions of trustworthy AI.

Breakthrough Results

AI4CCAM has already delivered a suite of innovative AI models addressing critical CCAM needs:

• Scene Explain – providing semantic and logical scene explanations based on structured metadata.
• VRU Trajectory Prediction– forecasting pedestrian and cyclist movements using social-force modelling.
• Scene Understanding – enabling precise segmentation of objects in the vehicle’s surroundings.
• Car Trajectory Prediction – predicting vehicle paths for improved situational awareness.
• VRU Crossing Time Prediction – estimating when pedestrians or cyclists are likely to cross, enhancing anticipatory safety.

All of these models bring explainability and foresight to the heart of automated driving.

Impact in Numbers

AI4CCAM’s progress speaks for itself:

• 3 VR experimentations and 5 AI models
• 23 scientific publications and 7 planned patent filings
• 13 logical scenarios for AI testing
• 1 open participatory platform fostering collaboration across sectors

The project also leaves a strong legacy of open science and collaboration, including an open-source model shared with the European Trustworthy AI Association, and a White Paper on Transparency as an Ethical Requirement, contributing to Europe’s leadership in responsible AI for mobility.

Towards a Trustworthy AI Future

AI4CCAM highlights that innovation in mobility is not only about automation, but also about building trust, ensuring that AI systems are transparent, ethical, and human-centric. By combining rigorous testing, stakeholder engagement, and scientific excellence, the project lays the groundwork for a future where AI-driven vehicles can safely and intelligently share the road with everyone.

To learn more about the project, please visit www.ai4ccam.eu and follow @AI4CCAM on LinkedIn for updates, publications, and access to its participatory space and media library.