Research On Local Motion Planning Of Self-driving Cars

Autonomous driving technology of the automobile can reduce traffic accidents caused by improper driving behavior,improve traffic efficiency and ensure traffic safety,so it has received extensive attention in recent years.As an important manifestation of the intelligent level of autonomous driving vehicles,the autonomous motion planning ability of automobiles has become the focus and difficulty of experts and scholars.Existing motion planning algorithms are mostly concentrated in the field of mobile robots.However,the external environment and the characteristics of the automobile are quite different from those of the robot.Therefore,it is very important to study the motion planning algorithms suitable for autonomous driving vehicles.In this paper,the autonomous vehicle is taken as the research object,and the local motion planning algorithm in two kinds of scenarios,unstructured road and structured closed road,is mainly studied,and the simulation experiment is designed to verify and evaluate the algorithm.The main research work of this paper is as follows:1.The constraints of vehicle motion planning are studied.Based on the reasonable analysis of vehicle motion state and the theoretical derivation of vehicle theory and geometric knowledge,vehicle kinematics constraints,lane boundary constraints,obstacle avoidance constraints and performance constraints of the algorithm for autonomous driving vehicle motion planning are established,which provide constraints for the development of local motion planning algorithm.2.The motion planning algorithm for unstructured road environment is studied.The commonly used map representation methods is analysed.Based on the characteristics of unstructured road environment,the global navigation layer and the local planning layer is used to describe the environment.The raster map is built in real time in the local planning layer with sensor information.The motion planning algorithm based on the A* algorithm is stydied.Since the traditional A* algorithm does not consider the vehicle outline,this paper adopts redundant safe space setting to avoid collision.Aiming at overcoming the shortcomings of the traditional A* algorithm,this paper improves the original heuristic function and uses the heading angle difference as the heuristic information.Compared with the original algorithm,it can significantly improve the smoothness of the path,so as to obtain a path which can satisfy the vehicle motion constraints.3.A local motion planning algorithm is developed for structured closed road environment.The typical scenarios under structured roads are constructed by using environment elements such as lane lines and relative positions of obstacles on structured roads.Each scenario is clustered.Combining with each scenario,a decision-making algorithm based on safety time threshold is proposed.Combining with pure tracking algorithm,a motion planning algorithm based on dynamic target points is proposed,which not only considers single-cycle trajectory planning,but also take into account the continuity of vehicle behavior decision-making,thereby conduciving to the applicability of the algorithm to actual dynamic traffic scenarios.4.The real vehicle experiment and multi-scene simulation experiments are designed to verify and evaluate the local motion planning algorithms of autonomous driving vehicles in different environments.The simulation traffic scene,sensor model and vehicle model are built by software named Pre Scan.Simulink modeling tool and C++ language are used to implement the algorithm.Several simulation experiments are carried out.The algorithm is evaluated and analyzed by the curvature of the path and the deviation of the path.The simulation results verify the validity of the local motion planning algorithm for autonomous driving vehicle developed in this paper in their respective application scenarios.A real vehicle experiment was designed for the structured road motion planning algorithm.The real vehicle experiment was completed in the real test site using Micro Autobox II tool.The real vehicle experiment proves that the decision planning algorithm developed in this paper also has strong robustness and real-time performance in the actual traffic scene.