Generating Autonomous Driving Test Scenarios based on OpenSCENARIO

Abstract

At present, with the rapid development of autonomous driving technology, simulation test has become an important means of autonomous driving technology verification and evaluation due to its high-test efficiency, strong repeatability, low cost, and process safety. However, the parameters in the test scenarios in the current simulation test are all fixed values, the test is only a discrete test, and a comprehensive test of the entire parameter space is lacking. Therefore, this paper proposes a generating method of autonomous driving test scenarios based on OpenSCENARIO and constructs a test scenario including parameter space. This method automatically builds the test scenario by constructing the configuration file of the test scenario and writing Python scripts. When executing the test scenario, the road elements, weather elements and traffic participant elements are used to screen the test scenarios that conform to the operation design domain of the autonomous driving, and the autonomous driving function is tested and verified through the Carla simulation platform. Finally, a test scenario is built for the lane keeping function. Experiments show that the test scenarios constructed automatically in this paper can not only meet the needs of autonomous driving simulation test, but also simplify the writing of test scenarios and greatly improve the efficiency of scenario construction.