Deliverable D3.3 of the AIthena project, ‘Report on final AI algorithm development’, showcases major advancements in artificial intelligence designed to make autonomous vehicles safer, more transparent and more trustworthy.
The report presents the final versions of state-of-the-art AI models that enhance autonomous vehicles’ ability to perceive their surroundings, predict risks, and make driving decisions. These developments focus on explainable AI (XAI), ensuring that automated driving systems can clearly communicate how and why they act, a crucial step towards gaining the public’s trust and ensuring safety.
Key innovations include:
- Explainable perception modules that reveal which visual features influence the car’s understanding of its surroundings.
- Advanced sensor fusion, which combines LiDAR, radar and camera inputs to enable more reliable object detection in challenging conditions.
- Safety-critical prediction models that can forecast collision risks and interpret complex traffic scenes using real and synthetic training data.
- Transparent decision-making algorithms enable vehicles to justify manoeuvres such as braking or steering in critical scenarios.
The report also updates AIthena’s machine learning lifecycle framework, improving standardisation, documentation and accountability in AI development by providing detailed model cards and MLOps tools.
These algorithmic breakthroughs are a significant step towards the deployment of autonomous mobility systems that users can trust. The results will be validated further in real and virtual test environments and shared through European data initiatives to accelerate research across the mobility sector.
The full AITHENA D3.3 report can be found at AITHENA-D3.3-Report-on-final-AI-algorithm-development.pdf