The core of this project is about the study of methods and tools for the simulation of autonomous driving systems, involving the modelling of sensors, the provision of a scenario library, the development of a method of analysis of the results simulations and the constitution of a methodological reference.
The project 3SA (Simulation for the Safety of Autonomous Vehicle Systems), which is an extension of the project SVA (Simulation for the safety of the Autonomous Vehicle), aims to go further in the use of digital simulation to demonstrate the safety of the individual autonomous vehicle.
The use of simulation for the demonstration of autonomous vehicle safety is an unavoidable alternative to physical validation tests. The main difficulty is to ensure that the vehicle behaves safely in the multitude of situations that it may face. The project 3SA, launched for four years, focuses on three major aspects of ensuring the safety of the autonomous vehicle (other than cybersecurity) :
- the dysfunctional aspects of systems and inter-systems notably related to software bugs or hardware failures,
- the functional safety (or SOTIF, Safety Of The Intended Functionality) centred on the erroneous decisions related in particular to a bad interpretation of the environment by the vehicle
- and the potentially predictable misuses of the driver with the aim of identifying and preventing misuses of the system.
In the continuity of the project SVA, we will pursue and delve deeper into the following activities: the modelling of sensors, the study of simulation methods and tools for autonomous driving systems and the construction of a library of potentially safe scenarios.
Other topics developed as part of 3SA:
- the construction of a methodological repository for validation plans and safety arguments,
- the definition of acceptance and validation criteria for a digital test,
- the quantification of the coverage of tests in relation to the situations that will be encountered by the vehicle in autonomous mode and of the part that is performed digitally,
- the demonstration of the representativeness of the simulation and scenarios.