This week we introduce next AIthena consortium partner: Valeo
Valeo is an independent group specialising in the development, production and sale of components, integrated systems and modules for passenger cars and commercial vehicles. Valeo is one of the world’s leading suppliers to the automotive industry. Today, Valeo is the world’s leading supplier of sensors for driver assistance. Since the 1980s, Valeo has been developing sensors for driver assistance in the field of ultrasound, radar, lidar, cameras and laser scanners.
Valeo maintains its own network of research sites (DAR – Driving Assistance Research) in Paris, Kronach and San Francisco, collaborates with universities and finances several doctoral theses.
What is the role of Valeo and the team in the project?
We are working on sensor fusion algorithms (lidar, camera) with the goal of improving the safety of Perception. Furthermore, we pursue the mitigation of the DNN insufficiency lack of generalization by the safety mechanism of weight fusion, which is also known under the terms model soup, model/weight averaging. This represents a practical alternative to deep ensembles, which require a lot of computing power and runtime at inference time.
In addition, we aim to publish our research on these topics in the form of papers at international conferences.
What are you currently working on in the project?
After conducting a detailed analysis on the state of the art of sensor fusion and weight fusion, preparatory work and first prototype developments are currently being carried out. Furthermore, we have worked with AIthena partners and organized the 1st AIthena practitioner’s workshop, which took place in conjunction with the consortium meeting in Eindhoven (15-16 November 2023). The aim of the workshop was the exchange on topics like the AIthena results of the user survey conducted in Work Package 1, and the concepts of Data and Model Cards and their use in the project.
From your perspective, how do you see the contribution of the AIthena project to building trustworthy, explainable, and accountable AI-based CCAM?
The development of trustworthy, explainable, and accountable Artificial Intelligence (AI) is of great importance for products where AI is in the safety-critical path, i.e. based on the decision of AI a safety hazard may arise. AIthena project offers the possibility to investigate the challenges within several components of the Autonomous Driving stack (perception, situational awareness and decision making) together with other partners. This provides the basis for a decisive impact in this area.
You can read more about Valeo at Valeo: 100 years of innovation.