Abstract: With the rapid development of autonomous driving technology, its key technologies can be divided into four modules: perception, decision, planning and control. Path planning plays an important role in them. A good path planning algorithm can plan an optimal path before automatic driving. In the process of driving, it can combine the environmental information sent back by perception and decision, and then plan a fine-tuned local path in real time. On the basis of completing functions such as obstacle avoidance and lane changing, the planning takes into account the safety and comfort of the driving trajectory. Based on Dijsktra shortest path planning algorithm, this paper starts with the discussion of global path planning and constructs a topological node map in a closed environment. Then improve the algorithm, which not applicable to the pain points in the real environment test of autonomous driving, so as to realize the global path planning task from any starting point to the end point. Finally, it is verified in the real vehicle test. The test results show that in a close and closed-loop environment, the self-driving vehicle can plan the ideal shortest path for driving.

Key words: Autonomous driving; Global path planning; Dijkstra algorithm; Closed environment construction

摘要： 随着自动驾驶技术的快速发展，其关键技术被划分为感知、决策、规划及控制四大模 块，路径规划在其中承担着重要角色。良好的路径规划算法能够在行车前规划出一条最优路 径，并在行车过程中能够结合感知、决策传送回来的环境信息，实时规划出微调后的局部路 径，在完成避障、换道等功能的基础上，兼顾着行驶轨迹的安全性、舒适性。文章从全局路 径规划出发，基于 Dijkstra 最短路径规划算法，构建封闭环境下拓扑节点地图，改进算法中 不适用自动驾驶实车测试中的痛点情况，以实现任意起始点至终点的全局路径规划任务。最 终在实车试验中进行验证，试验结果表明，在封闭且闭环的环境下，自动驾驶汽车可规划出 理想的最短路径进行行驶。

关键词: 自动驾驶；全局路径规划；迪杰斯特拉算法；封闭环境建模