基于点云协方差描述子的多机器人目标识别与编队跟踪

发布时间：2018-11-08 19:44

【摘要】：针对多移动机器人目标识别及编队跟踪问题,提出一种基于点云协方差描述子的目标识别方法及编队跟踪控制方法.为降低机器人端机载处理器负荷,基于Web Socket协议搭建网络架构.在此基础上,通过引入点云协方差描述子进行目标检测,并利用支持向量机完成离线建模.结合支持向量机分类器与Kullback-Leibler Divergence(KLD)-Sampling自适应粒子滤波算法,实现目标部分遮挡下的在线识别,得到目标点云跟踪位置信息.利用势场函数和有向刚性图论综合设计编队跟踪控制器,实现多机器人编队目标跟踪.最后通过实物平台进行实验,结果表明,所提出的基于点云协方差描述子的多机器人目标识别与编队跟踪算法,与传统方法相比,跟踪收敛时间缩短4,s,跟踪精度提高约2.5%,,通过搭载有限数量传感器,可以更有效地解决多机器人编队目标跟踪问题.
[Abstract]:Aiming at the problem of target recognition and formation tracking of multi-mobile robots, a method of target recognition and formation tracking control based on point cloud covariance descriptor is proposed. In order to reduce the load of airborne processor on the robot side, the network architecture based on Web Socket protocol is built. On this basis, the point cloud covariance descriptor is introduced to detect the target, and support vector machine is used to complete the offline modeling. Combined with support vector machine classifier and Kullback-Leibler Divergence (KLD) Sampling adaptive particle filter algorithm, the online recognition under partial occlusion of target is realized, and the location information of target point cloud tracking is obtained. The formation tracking controller is designed by using potential field function and directed rigid graph theory to realize multi-robot formation target tracking. Finally, the experiments are carried out on the physical platform. The results show that the proposed multi-robot target recognition and formation tracking algorithm based on point cloud covariance descriptor can shorten the tracking convergence time by 4 s compared with the traditional method. The tracking accuracy is improved about 2.5%, and the problem of multi-robot formation target tracking can be solved more effectively by using a limited number of sensors.
【作者单位】：天津大学电气自动化与信息工程学院;
【基金】：国家自然科学基金资助项目(61673294,61273092)~~
【分类号】：TP242

【相似文献】