A Novel Simulation Framework for the Design and Testing of Advanced Driver Assistance Systems

摘要

The number and complexity of newly developed automated driving systems has been constantly rising over the past decade. Especially the introduction of vehicle-to-everything (V2X) communication is expected to further potentiate this development. In order to be deployed, the functional safety of the developed systems has to be assured previously. However, the testing in a representative number of field tests is costly and time-consuming. For this reason, virtual test drives have risen as an important option for design and testing of automated driving technologies, leaving only the final validation to test with real vehicles and thus significantly reducing the overall expenditure. The authors of the work at hand introduce a simulation framework based on the vehicle simulator CarMaker, complemented with the middleware platform Robot Operating System (ROS) and fed with real traffic data, which allows to automatically test advanced driver assistance systems for a large number of real world scenarios by varying topology, vehicle and communication parameters, among others. The simulation framework is then used to demonstrate the benefit of collective perception (i.e. sharing of on-board sensor data among nearby vehicles by V2X communication) for a vehicle merging into a freeway, with metrics such as the vehicle awareness on spot and the time it has to plan and execute its maneuver.