Virtual Co-Simulation Platform for Test and Validation of ADAS and Autonomous Driving 2019-01-5040
Vehicles equipped with one or several functions of Advanced Driver Assistant System (ADAS) and autonomous driving (AD) technology are more mature and prevalent nowadays. Vehicles being smarter and driving being easier is an unstoppable trend. In the near future, intelligent vehicles will be mass produced and running on the road. However, before the mass-production of intelligent vehicles, a lot of experimental tests and validations need to be carried out to insure the safety and reliability of ADAS and AD technology. Although the road test of real vehicles is the most reliable and accurate test method, it cannot meet the need of rapid development of technology research due to high time and financial cost. Therefore, a high-efficient design and evaluation methodology for ADAS and AD development and test is a must. In this paper, a virtual co-simulation platform based on MATLAB/Simulink, OpenModelica and Unity 3D game engine (MOMU) is proposed. Simulink is used for vehicle control software modeling. OpenModelica simulates vehicle dynamic models written in Modelica. Unity provides visualized display of ADAS simulation and a virtual city model built in Unity containing sensor models, vehicle models and road environment models. Furthermore, Machine Learning Agents Toolkit (ML-Agents), an open source toolkit, integrated in Unity allows users to control the Agent in Unity through Python API. Because of this feature, the platform MOMU can test and validate not only ADAS but also AD technology in the phase of design. In this paper, two cases concerning Autonomous Emergency Braking system (AEB) and autonomous driving of an electric vehicle are designed to illustrate the establishment of MOMU.