A Multi-Dimensional Evaluation Model for Autonomous Vehicles Based on Choquet Integral

The testing and evaluation technology for autonomous vehicles (AVs) is a crucial aspect supporting the development of vehicle electrification and intelligentization. Break-throughs in core technologies are urgently needed. In terms of intelligence evaluation, this paper constructs a behavior performance evaluation model for autonomous vehicle intelligence using complexity science and information measurement theory, based on the input conditions of the tests. Considering the completion status of the tests, the model integrates machine learning, Turing tests, multiple experiments, and natural driving data to establish a comprehensive subjective and objective evaluation theory. This theory encompasses driving autonomy, learning evolution, and traffic coordination. On this basis, a behavior performance model based on the Choquet Integral is developed to determine the weight of each evaluation indicator in the overall assessment score, enabling a quantitative evaluation of intelligent behavior performance for various driving tasks across different driving scenarios.