基于优化Hector-SLAM算法的机器人自主导航系统设计

Design of robot autonomous navigation system based on optimized Hector-SLAM algorithm

针对机器人在复杂地形环境下定位不准确、建图精度不高和误差堆积等问题,设计了一种基于优化Hector-SLAM(simultaneous localization and mapping,实时定位与建图)算法的自主导航系统,以实现机器人在及时躲避障碍物的同时准确到达目标点。首先,采用双三次插值法代替原有插值法,以解决Hector-SLAM算法在使用低精度激光雷达数据建图时易出现地图不清晰等问题,从而提升扫描匹配的精准度。其次,采用扩展卡尔曼滤波算法对里程计和惯性测量单元的数据进行融合,以提高定位的准确性。再次,针对激光点数据无法瞬时获得而机器人持续运动所导致的运动畸变问题,结合里程计辅助法与PL-ICP(point to line iterative closest points,点到线迭代最近点)配准法,以实现运动畸变的校正;同时,设置倾斜角阈值以消除地图重影,并利用A*算法和动态窗口法规划最优路径。最后,以AGV(automated guided vehicle,自主导引车)为例,在实际场景中开展建图实验和自主导航实验。结果表明,优化后自主导航系统的平均建图相对误差约为0.44%,最小建图误差为0.236 m,较优化前减小了0.041 m,有效地解决了因误差堆积和运动畸变而导致的建图不清晰问题,增强了AGV在复杂地形环境中的适应能力,实现了高精度定位。研究结果对提高移动机器人在室内多障碍物环境下的自主导航能力具有一定的理论和工程意义。

Aiming at the problems of inaccurate positioning,low mapping accuracy and error accumulation of robots in complex terrain environments,an autonomous navigation system based on optimized Hector-SLAM(simultaneous localization and mapping)algorithm is designed,which ensures that the robot reaches the target point accurately while avoiding obstacles promptly.Firstly,the bicubic interpolation method was used to replace the original interpolation method to solve the problem of map blurring when the Hector-SLAM algorithm was used to build maps with low-precision lidar data,so as to improve the accuracy of scan matching.Secondly,the extended Kalman filter algorithm was used to fuse the data of odometer and inertial measurement unit,which could improve the positioning accuracy.Thirdly,in view of the problem of motion distortion caused by the inability to obtain instantaneous laser point data and the continuous movement of robot,the odometer auxiliary method and PL-ICP(point to line iterative closest points)registration method were combined to realize the correction of motion distortion.At the same time,the tilt angle threshold was set to eliminate the map ghosting,and the optimal path was planned by using A*algorithm and dynamic window approach.Finally,taking AGV(automated guided vehicle)as an example,the mapping experiments and autonomous navigation experiments were carried out in actual scenarios.The results showed that the average mapping relative error of the optimized autonomous navigation system was about 0.44%,and the minimum mapping error was 0.236 m,which was 0.041 m less than that before optimization.It effectively solved the problem of unclear mapping caused by error accumulation and motion distortion,and enhanced the adaptability of AGV in complex terrain environment,so as to achieve high-precision positioning.The research results have certain theoretical and engineering significance for improving the autonomous navigation ability of mobile robots in indoor multi-obstacle environment.