Data-Driven Vehicle Cut-In Test Cases Generation for Testing of Autonomous Driving on Highway

In order to generate test cases covering complex and varied real-life traffic in a scenario-based test, multiple real samples of vehicle cut-in scenarios were extracted from highD dataset. A description model of vehicle cut-in scenario was built based on analysis of the positional relation between motion parameters and participated vehicles from these real samples. The risk degree of this scenario was evaluated depending on TTC in cut-in point. Finally, test cases generation was executed using Monte Carlo method by combining with the distributions of parameters in description model. The results show that the generated cut-in test cases are capable of covering all risk levels, which can better support autonomous driving testing.