An objective evaluation method for automated vehicle virtual test

Abstract

Reasonable test scenarios and objective evaluation methods could rapidly promote the development of autonomous technology. The current quantitative evaluation method of automated vehicles could comprehensively evaluate the performance of autonomous driving systems rather than only focused on effectiveness. However, it needed massive participation of experts, which led to the strong subjectivity for the evaluation result. In order to reduce the participation of experts and implement automatic test evaluation, especially in the virtual environment, an objective automated vehicle evaluation method was proposed. The proposed methodology utilized the idea of Multi-Criteria Decision Making (MCDM). It combines the analytic hierarchy process (AHP) method and the improved criteria importance though intercriteria correlation (CRITIC) method to determine the weight of the evaluation index. Then the adaptive grey relational analysis (GRA) algorithm was implemented to evaluate the performance of the automated vehicle. The experiment in a virtual environment scenario showed that the evaluation results of the proposed method had high consistency with the traditional human intervention evaluation algorithm. Meanwhile, it reduced the subjectivity and saved the cost of inviting a large number of experts to participate in the evaluation.