Autonomous Vehicle Test Scenario Construction Method Based On Hazard Analysis

Autonomous vehicles(AVs)have profound impact on reducing traffic congestion,accident rates,and helping cities build safe and efficient future travel structures.However,before their large-scale entry into the market,it is necessary to conduct billions of kilometers of road testing to verify the feasibility and safety.Even if this was possible,it would not cover the unknown scenarios.The development of virtual test technology provides new ideas for AVs testing.The test scenarios serve as the basis for testing and play a vital role in AVs testing.Existing test scenario construction methods exist problems such as only known scenarios can be constructed,the number of scenarios is large,and long test period are required.Therefore,this paper proposes an AVs scenarios construction algorithm based on hazard analysis technology,covering unknown test scenarios as much as possible,and achieving efficient and reliable testing.From the perspective of AVs’ danger,the influencing factors of AVs accidents are analyzed,and the probability density function is used to describe the characteristics of them.Then the FMECA is used to extract following,avoidance and lane changing scenarios as dangerous scenarios.At the same time,the STPA is used to analyze the dangerous behaviors of AVs,and extract AVs safety constraints.On this basis,combined with FTA to carry out top-down specific designs for different dangerous scenarios,so the combination of elements in different dangerous scenarios are obtained.And the association conditions between its elements for different dangerous scenarios is designed to provide a basis for the dangerous scenario construction.The collision situation of specific following,avoidance and lane changing scenarios are analyzed,and different dangerous boundary models based on collision conditions are obtained respectively.The characteristics,combinations and relationships of the elements are combined,then all the conditions for generating the specific scenarios are formed.Due to the continuity of the function,the dangerous test scenarios are infinite.So,the optimization theory based on constraints is used to construct representative specific test scenarios closing to the dangerous boundary by taking the dangerous boundary model as the objective function and the characteristics,combinations and relationships of the elements as the feasible region.Constructed representative scenarios verifies the algorithm feasibility.And the paper evaluates the algorithm reliability from the aspects of performance and effect.Finally,the application of the algorithm is explained.