Yann Duplouy joined SystemX in 2015 to work on a thesis dedicated to the application of formal methods to the control of self-driving cars. He revisits his thesis that he completed within the SVA (Simulation for the Autonomous Vehicles Safety) project of SystemX and supervised by ENS Paris-Saclay.
What was the topic of your PhD thesis?
My thesis focused on the application of formal methods to the control of the autonomous vehicle.
I was interested in this topic because the creation of systems that allow vehicles to be autonomous in certain situations is a key challenge for all players in the automotive industry. Nevertheless, the modelling of vehicles and their environment is complex, and often requires a long time for computation. Instead of using a complete physical simulation (or directly running the vehicles), exact but very time-consuming techniques to produce the expected results, some industry players decide to cut the bigger problem into smaller problems, testing for example the wear and tear of a shock absorber or the aerodynamics of the bodywork. Among these problems is the central question of the decisions taken by the onboard controller in these vehicles. As part of my thesis, I developed a method to directly test the onboard decision controller in a sufficiently optimal way to meet the standards set. The onboard decision controller designates an algorithm or a decision procedure, inside the vehicle, which will choose the behaviour to be applied (for example an emergency brake if an obstacle is found on the road).
What do you remember from your doctorate?
I had done several academic internships before going to SystemX. The difference between such internships and a thesis is that, with a total of three years, you have enough time to perform an extensive study of the state-of-the-art of the domain. Moreover, doing the thesis at SystemX gave me an insight on how different industrial research can be from academic research. I was fortunate to be able to make a real contribution, on both academic and industrial point of views, in the field of the application of statistical verification to the analysis of performance of autonomous vehicles.
Can you tell us your favourite memory from your time with SystemX?
SystemX gave me access to renowned industry conferences. I remember in particular the conference of the Society of Automobile Engineers in 2017, during which I had the opportunity to meet key players of the autonomous vehicle industry and to present my thesis work.
What are your plans for the future?
I just started working in a post-doctorate position at the Loria laboratory (Lorraine research laboratory in computer science and its applications) at Inria Nancy Grand Est. My research project focuses on the application of statistical verification methods to the analysis of distributed programs.
Read more about Yann Duplouy
Name of the PhD thesis: Application of formal methods to the control of self-driving cars
Qualification obtained before the doctorate: Master of research in computer science – Université Paris-Saclay, Université Paris Diderot, ENS Paris
R&D project: Simulation for Autonomous Vehicle Safety (SVA)