Skip to main content
search results
Sorry, but nothing matched your search terms. Sorry, but nothing matched your search terms. Sorry, but nothing matched your search terms. Désolé, mais rien ne correspond à vos critères de recherche. Désolé, mais rien ne correspond à vos critères de recherche. Entschuldigung, wir haben nichts zu diesem Suchbegriff gefunden.
Sorry, but we cannot handle your search query now. Please, try again later! Sorry, but we cannot handle your search query now. Please, try again later! Sorry, but we cannot handle your search query now. Please, try again later! Désolé, mais nous ne pouvons pas traiter votre demande. Veuillez réessayer plus tard ! Désolé, mais nous ne pouvons pas traiter votre demande. Veuillez réessayer plus tard ! Entschuldigung, wir können Ihre Suchanfrage zurzeit nicht bearbeiten. Bitte versuchen Sie es später noch einmal.
Search suggestions

Autonomous Vehicles: Expleo partnered with IRT SystemX to evaluate and improve the performance of AI decision-making systems

Paris, 27 January 2020. Expleo, the technology partner for innovative companies, announces its participation, alongside Apsys, PSA Group and Naval Group, in the EPI research project (AI-based Decision Making Systems’ Performance Evaluation).
Scroll

Paris, 27 January 2020. Expleo, the technology partner for innovative companies, announces its participation, alongside Apsys, PSA Group and Naval Group, in the EPI research project (AI-based Decision Making Systems’ Performance Evaluation), led by IRT1 SystemX, an accelerator for the digital transformation of industry, services and territories. Launched in 2018 for a three-year period, this project aims to develop a methodology for the validation of decision-making systems with embedded Machine Learning in the field of autonomous vehicles and maritime transport.

The EPI project particularly focuses on developing the autonomous vehicle. One of the main technological obstacles to overcome lies in the performance evaluation and optimisation of embedded Artificial Intelligence (AI) used in decision-making. 

Although intelligent solutions already exist for obstacle detection and avoidance (optical cameras, radar, sonar and lidar technology, etc.), the ability of an autonomous vehicle to adopt a safe and consistent behavior, via artificial neural network systems, remains to be proven.

A vehicle’s perception and interpretation of its environment could be hampered by unforeseen events (exceptional weather conditions, malfunction of the vehicle’s optical sensors, etc.), that could lead to errors in assessment and decision-making (inability to detect traffic signs or a pedestrian crossing in front of the vehicle, etc.). Research carried out under the EPI project aims to optimise the stability and robustness of the AI ​​system guaranteeing the quality of its decision-making. 

In partnership with SystemX, Expleo created a report on different existing AI evaluation techniques. Expleo is also involved in developing new techniques for assessing AI robustness. The company participated in improving the robustness of artificial neural networks by introducing 3D disturbances (vibrations, camera movements, etc.) into the input data. 

Dimitri Bettebghor, Chief Data Scientist at Expleo explains,

“With the EPI project, the industry is exploring the potential of AI, a key issue for the future of mobility. The question of an autonomous decision-making function confronts us with real challenges in terms of the validation and certification of critical systems. This is why collaborative and multi-disciplinary research is the most adequate model for advancing within an area of this scope. Expleo is very proud to participate, alongside IRT SystemX and all of its industrial partners, in this cutting-edge technological research project. ”

The EPI project also focuses on estimating the level of machine learning. By applying an empirical learning method, an AI-based decision-making system proceeds by systematically comparing what it “sees” with what it has already seen or knows. In the current state of technology, the performance of these algorithms is based on their exposure to a wide variety of driving and environmental scenarios, the proper number and distribution of which must be precisely assessed.  

If the EPI project aims to achieve two Proofs of Concept in the automobile and maritime sectors, its scope could extend to many others (the military, medicine, aeronautics, etc.). By working on the development of AI-specific standards, this project could lead to the certification of critical AI-based systems in France. 

Media contacts 

Expleo  
Valentine Piedelievre-Eman 
valentine.eman@expleogroup.com 
+33 (0)6 07 27 68 55  

L’été en hiver  
Morgane Gens  
morgane.gens@eteenhiver.com   
+33 (0)6 88 63 34 30  

Download

Download whitepaper