Evaluating Autonomous Functions Performance Through Simulation Using Interoperable Sensor, Vehicle, and Environment Models

Abstract

The evaluation, validation, certification, and homologation of connected and automated autonomous vehicles will require more and more virtual testing because of a new level of system complexity the automotive industry has never faced before. Ensuring the robustness, accuracy, and reliability of the perception stages of an "intelligent" and automated vehicle is recognized as one of the challenges for the transition of future mobility means to the higher levels of autonomous driving. From a technological perspective, safety is enforced using combinations of sensors systems exploiting different physics principles used by GPS, by optical sensors, by near or thermal infrared sensors, by electromagnetic sensors and by ultrasonic sensors. The consistency of the performance verification, expected for the embedded processing relying on artificial intelligence software stacks, depends on the driving environment's characteristics. It is now obvious that real testing should evolve accordingly and should be completed by virtual testing. Therefore, the industry has embraced the idea of introducing simulation methods to support such validation. At ESI and with our strategic research and development partner University Gustave Eiffel (merge between previously IFSTTAR and 5 other organisms) we are working on the development of new interoperable and interconnected simulation solutions offering the proper agility related to the integration of new methods. In the new generations of simulation platform, the problem is to address a complex multi-layer simulation task not only processed by one software and one computer but by several interoperable remote platforms (i.e Dynamics vehicle modelling, sensors modelling, communication modelling, and traffic modelling). Such a simulation platform must respect important constraints such as the interconnection with real-time data management and processing software, or the capability to share the simulation components, the processing stages, and the driving functions across multiple applications and/or computers. Simulation can evaluate the performance of complex systems from well-defined scenarios and configurations. It intervenes in several phases of system manufacturing in the design and exploitation phases. With virtual prototyping, simulation can be used to predict in real time the system behavior with respect to its control in order to implement optimal strategies. It is also used for verification, validation and improvement of choices in the design phases. Our technical paper aims to present what we did with partners and customers in order to confirm the value brought by such new simulation solution.