Graphical Task Modeling Method of Autonomous Vehicles for Specific Scenes

Specific scenes featured by relatively fixed scenario and rich operational tasks is the main application area of autonomous vehicles, and task modeling expression can improve the generality and scalability of the autonomous driving system by providing standard expression of autonomous driving tasks in specific scenes. To improve the development efficiency of task modeling, a graphical task modeling method of autonomous vehicles for specific scenes is proposed. Firstly, a graphical task modeling system architecture based on high-precision maps is presented. Secondly, according to the characteristics of the operational tasks in specific scenes, the graphical expression of the task models is designed to construct the operational tasks graphically based on the map library and task library, and the operational scheme of the autonomous vehicle associated with the scene is formed according to the graphical task flow. Then the operational scheme is verified with the rule verification method of the task flow based on the rule library to generate the corresponding operation scheme file. Finally, the generated operation scheme file is imported to the simulation platform for testing and verification. The experimental results show that compared with the text-based modeling method, the method can construct the autonomous vehicle operation task flow under specific scenes intuitively and accurately, reduce the difficulty of task modeling, and improve the development efficiency of the autonomous driving system.