Abstract: Aiming at the problem that robot operating system (ROS)-based unmanned vehicles in the university student intelligent vehicle competition are prone to environmental interference and fail to achieve autonomous navigation, this paper proposes a lidar-based autonomous navigation algorithm for traffic cone detection and active avoidance. First, the DBSCAN clustering method is employed to process point cloud data within regions of interest, enabling precise identification of traffic cones' spatial coordinates. Subsequently, the Delaunay triangulation algorithm is applied to the clustered cone center points to establish topological relationships, followed by path smoothing using cubic spline interpolation combined with weighted averaging of current and historical data points. Finally, the planned trajectory is transmitted via ROS topic messages to the vehicle's low-level control module for autonomous tracking. Experimental validation conducted in outdoor simulated competition environments demonstrates that the proposed navigation algorithm meets competition requirements, effectively preventing track deviation caused by external interference. This work provides a valuable reference for designing autonomous navigation systems for ROS-based unmanned vehicles in university student intelligent vehicle competitions.

Key words: university student intelligent vehicle competition; LiDAR; autonomous navigation; point cloud clustering

摘要： 针对大学生智能汽车竞赛中机器人操作系统（ROS）无人小车易受到外界环境干扰无 法完成自主导航的问题，文章提出了一种依靠激光雷达检测锥桶并实现主动避让的自主导航 算法。首先，使用 DBSCAN 聚类方法对感兴趣区域的点云进行聚类，得到感兴趣区域内锥桶 的空间位置坐标；其次，使用 Delaunay 算法对聚类后的锥桶中心位置点集进行三角剖分，利 用三次样条插值方法并通过对当前点和历史点的加权平均来平滑路径；最后，将得到的规划 路径通过 ROS 话题消息发送到小车的底层控制模块来控制小车跟踪行驶，并在室外模仿赛场 环境下进行实验验证。结果表明，设计的自主导航算法能够满足比赛的要求，避免了室外的 其他干扰造成小车行驶偏离赛道的情况，为大学生智能汽车竞赛中 ROS 无人小车的自主导航 系统的设计提供了参考。

关键词: 大学生智能汽车竞赛；激光雷达；自主导航；点云聚类