ASTRA, the Inria/Valeo project team working on tomorrow's smart mobility solutions

Fawzi Nashashibi (F.N.): The idea for the ASTRA project originated at Inria, as part of our desire to develop strategic partnerships in the fields of autonomous vehicles and smart mobility. After previous collaborations with Valeo, we considered it natural to offer Valeo the opportunity to become our partner in this joint research project.

Benazouz Bradai (B.B.): We have been working regularly with Inria for almost 15 years now, via theses and joint or collaborative projects. But until now, it has been as part of a more opportunistic, collaborative approach: “I have a specific subject. Inria has expertise in this field, and we are launching a specific contract or thesis to make headway on this subject". The ASTRA project has enabled the definition of guidelines and the implementation of common resources. 

F.N.: The joint team was created in February 2022. The initiative has received support from the France Relance recovery plan, which aims to promote industrial innovation, particularly for autonomous vehicles, and a financial contribution from the French National Research Agency.

F.N. : The four-year ASTRA project focuses on smart terrestrial mobility, and specifically on highly automated and connected vehicles. This includes technologies such as computer vision, location, artificial intelligence, decision-making systems and vehicle-infrastructure communication. The aim is to develop reliable and deployable systems for autonomous vehicles.

B.B.: The scientific and technical objective is to develop methodologies and architectures for highly automated vehicles, beyond level 2. We are aiming for a more advanced level of automation, involving high-performance, certifiable and deployable concepts and are steadily gaining momentum. 2025 will be an important year. We can now start to integrate our developments into small demonstrators. 

F.N. : We are focusing on four main areas. 

1. Perception and interpretation of the environment: object detection, lane detection, fusion of data from different sensors, semantic segmentation of scenes, and data processing in poor weather conditions or low visibility.
2. Localisation, with a focus on improving the accuracy and integrity of location by merging data from the vehicle's sensors with that from GNSS systems (geolocation and navigation by satellite system), or to replace them.
3. Decision-making: this includes developing algorithms to plan vehicle trajectories and manoeuvres.
4. And lastly, the pooling of data for smart mobility: including the use of data collected by several vehicles to improve location, mapping and driving decisions, particularly in dense traffic, driving in close proximity, etc.

F.N. : The joint team is composed of around 30 people, 12 from Valeo (six from Valeo anSWer and six from Valeo.ai) and 15 or so from Inria. This collaboration is based on complementary approaches: on the one hand, Inria's strongly upstream research-based approach, with a very long-term vision, and on the other, Valeo's contribution with its key expertise in practical applications in an industrial environment, development, implementation, compliance with standards, certified software architectures, and so on. 

B.B. : Valeo contributes its pragmatic vision as a manufacturer and key player in the automotive industry, and as a leader in mobility. We have obtained an overview of the market and its development, thanks to our close contacts with all the manufacturers. We know the trends, we know where we're going in 5 years' time, in 10 years' time... We know what is and isn't feasible in terms of integration...

F.N. : The team has already made significant progress in a number of areas, such as perception and decision-making, by using a single sensor to reconstruct 3D environments, for example, which we consider to be a resounding success. 

B.B. : Progress has been made in decision making, especially through work on trajectory prediction, described in a publication which won the Georges Sarridis Prize for “Best Paper”. The team has also made significant progress in generating autonomous vehicle manoeuvres, which are currently being integrated and tested in simulation, with integration and testing on real prototypes coming soon. 

B.B. : The teams at both Inria and Valeo have clearly grown in stature as the project has progressed. At Valeo, our engineers are harnessing new ideas arising from fundamental research and out-of-the-box thinking. This creates powerful synergies and nurtures high-quality scientific discussions.

F.N. : At Inria, what we have gained is the manufacturer’s perspective. The desire to obtain not just theoretical metrics, but an actual system, right through to its implementation, integration and certification.

B.B. : A number of challenges remain, especially concerning perception. How can we guarantee the robustness of the algorithms in degraded situations, such as in rainy conditions or at night? The generalisation of algorithms to scenarios not seen during the training phase is a major challenge. Another challenge is to ensure the explainability of artificial-intelligence-based algorithms, which is essential for ensuring the safety of autonomous vehicles.

F.N. : Real-time map updating and the management of unmapped situations are also major challenges. The aim is to enable vehicles to react reliably to dynamic environments and guarantee a high degree of safety.

Valeo, a technology-based company and partner to all motor vehicle manufacturers and new mobility players, is committed to innovating for cleaner, safer and smarter mobility. Valeo is a technological and industrial leader in electrification, driving aids, the reinvention of life on board, and lighting inside and outside the vehicle. These four fields, which are key to the transformation of mobility, are the Group's growth vectors.

Key figures for Valeo: €22 billion in revenue in 2023 | 109,600 employees, 28 countries, 159 production sites, 64 research and development centres and 19 distribution platforms at 30 June 2024.

Valeo is listed on the Paris stock exchange.

Valeo website: https://www.valeo.com/
Watch the interview on Valeo's website: “ASTRA, the Inria/Valeo joint project, moves up a gear in 2025”