Generalization Method of Virtual Test Scenarios for Autonomous Vehicles

Automatic driving is an inevitable development trend in the traffic field. However, unlike traditional driving modes, lots of simulation test scenarios for autonomous vehicles are required before the operation in real environment to ensure the safety during the actual operation. At present, the actual environment testing method has high cost and low efficiency and cannot test a large number of test scenarios within a certain time period. Therefore, the paper mainly studies the generalization method of virtual test cases for autonomous vehicles. The test scenarios are divided into three types: functional scenario, logical scenario, and concrete scenario. The main indicators of the actual test environment and their hierarchical relationship are analyzed, and the indicator hierarchical model is established. Based on the hierarchical relationship of different indicators and multi-tree building method, the paper studies the generalization method of functional scenarios. Then, the distribution functions and values’ range of parameters are obtained by extracting and analyzing the quantifiable parameters of function scenario, and the logical scenarios are obtained. Within the range of different indicators, the selection method of specific values is studied. By analyzing the spatial relationship between different parameters, the combination method between these parameters is determined, and the concrete scenarios that can be used in actual test cases are obtained. Finally, the feasibility of the proposed method is verified by a case study.