Driving Evaluation System in Specific Scene

Test evaluation is an important procedure in development process of autonomous vehicle, and efficient evaluation system can greatly improve the development efficiency of autonomous driving system. Currently, autonomous driving in specific scenes has become the primary application scenes. To satisfy the requirement of autonomous driving in specific scenes, a scenario-specific autonomous driving simulation evaluation system is proposed. Firstly, the framework of scenario-specific autonomous driving evaluation system is put forward based on the entropy value model and sequential relationship analysis. The elements of tasks performed by autonomous vehicles are extracted and their weights are calculated. Then the task elements are ranked and combined weight of the task is calculated based on the weight of each element. Finally, the evaluation results are calculated based on the completion of each task element. Experimental results show that the proposed method can evaluate the completion quality of autonomous vehicle tasks for specific scenarios comprehensively and accurately, and greatly reduce the development cost of autonomous driving system for specific scenarios.