作者：包加桐1,2, 杨圣奥1, 朱润辰1, 唐鸿儒1, 宋爱国2
作者单位：1. 扬州大学电气与能源动力工程学院，江苏 扬州 225300;
2. 东南大学仪器科学与工程学院，江苏 南京 210096
关键词：重定位;机器人导航;视觉定位;粒子滤波
摘要：
针对室内移动机器人初始位姿估计和“绑架”问题，提出一种视觉与激光结合的重定位方法。预先开展基于视觉与基于激光的同步定位与建图，记录相机位姿和机器人位姿并生成位姿映射表，生成视觉稀疏特征点地图与二维占用栅格地图。重定位时，基于图像特征点匹配与高效透视N点算法估计视觉地图中的全局位姿。根据最近邻匹配从位姿映射表中检索出最佳的栅格地图中的全局位姿，完成基于视觉的粗定位。该位姿进一步作为自适应蒙特卡罗定位算法中的初始位姿并在其周围分布粒子。利用运动信息和观测信息更新粒子直至收敛，完成基于粒子滤波的精定位。与其他重定位方法进行比较实验，实验结果表明提出的方法定位准确、时间短且成功率高。
Relocation method for indoor mobile robot by combining vision and laser
BAO Jiatong1,2, YANG Shengao1, ZHU Runchen1, TANG Hongru1, SONG Aiguo2
1. College of Electrical Energy and Power Engineering, Yangzhou University, Yangzhou 225300, China;
2. School of Instrument Science and Engineering, Southeast University, Nanjing 210096, China
Abstract: In order to address the problems of initial posture estimation and kidnapping for indoor mobile robots, a relocation method by combing vision and laser is proposed. The vision-based and laser-based simultaneous location and mappings (SLAMs) are performed in advance. The camera postures and robot postures are recorded accordingly, resulting in a posture mapping table. The visual sparse feature point map and 2D occupancy grid map are generated. When relocation is started, the global posture within the vision map is estimated based on the image feature point matching and the efficient perspective-n-point (EPnP) algorithm. The best global posture within the grid map is retrieved through nearest neighbor matching while the vision-based coarse location is finished. The posture is further served as the initial posture around which the particles are distributed for the adaptive Monte Carlo location (AMCL) algorithm. The information about motion and observation is employed to update the particles until their convergence and the particle filter based fine location is achieved. The proposed method is compared to other relocation methods and the experimental results show that it has the advantage of accurate positioning, short time and high success rate.
Keywords: relocation;robot navigation;vision-based localization;particle filter
2021, 47(11):1-7,20 收稿日期: 2021-06-29;收到修改稿日期: 2021-07-28
基金项目: 国家自然科学基金资助项目(61806175)；扬州大学2020年校“青蓝工程”资助项目(2020YZUB03)
作者简介: 包加桐（1983-），男，江苏兴化市人，副教授，博士，研究方向为机器人传感与控制
参考文献
[1] LIM J H, LEONARD J J. Mobile robot relocation from echolocation constraints[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(9): 1035-1041
[2] 曹风魁, 庄 严, 闫 飞, 等. 移动机器人长期自主环境适应研究进展和展望[J]. 自动化学报, 2020, 46(2): 205-221
[3] 李昌明, 梅莉, 秦东兴. 基于扩展卡尔曼滤波的轮式移动机器人定位技术[J]. 中国测试, 2011, 37(6): 76-79
[4] THRUN S, FOX D, BURGARD W, et al. Robust Monte Carlo localization for mobile robots[J]. Artificial Intelligence, 2001, 128(1–2): 99-141
[5] HESS W, KOHLER D, RAPP H, et al. Real-time loop closure in 2D LIDAR SLAM[C]//2016 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2016: 1271-1278.
[6] YU S K, YAN F, ZHUANG Y, et al. A deep-learning-based strategy for kidnapped robot problem in similar indoor environment[J]. Journal of Intelligent & Robotic Systems, 2020, 100(3-4): 765-775
[7] CAO J, ZENG B, LIU J Q, et al. A novel relocation method for simultaneous localization and mapping based on deep learning algorithm[J]. Computers & Electrical Engineering, 2017(63): 79-90
[8] 黄烨笙, 徐郑攀, 陈远明, 等. 无人艇自主靠泊控制系统设计[J]. 中国测试, 2020, 46(10): 111-117
[9] MUR-ARTAL R, TARDOS J D. ORB-SLAM2: An open-source slam system for monocular, stereo, and RGB-D cameras[J]. IEEE Transactions on Robotics, 2017, 33(5): 1255-1262
[10] 蒋林, 向超, 朱建阳, 等. 加载语义似然估计的粒子滤波重定位[J]. 电子学报, 2021, 49(2): 306-314
[11] RUBLEE E, RADAUD V, KONOLIGE K, et al. ORB: An efficient alternative to SIFT or SURF[C]//2011 IEEE International Conference on Computer Vision, 2011: 2564-2571.
[12] GRISETTI G, STACHNISS C, BURGARD W. Improved techniques for grid mapping with Rao-Blackwellized particle filters[J]. IEEE Transactions on Robotics, 2007, 23(1): 34-46
