基于信息增益一致性的多机器人地图融合算法

Multi-Robot Map Merging Based on the Consistency of Information Gain

针对多机器人协同SLAM(同步定位与地图构建)的地图融合中,由于通信距离受限或网络拓扑变化造成信息缺失、从而影响全局地图构建的问题,提出一种基于信息增益一致性原理的动态地图融合算法.该算法是完全分布式的,且不依赖于任何特殊的机器人通信网络结构.该算法利用机器人所测局部地图的历史数据和当前数据之间的新增信息,使每个机器人都能同步地获取一致的、最新的全局地图.在有限的网络连接条件下,所提出的地图融合算法能够通过渐近收敛的方式获得准确的全局地图.在每一次迭代中,每个机器人得到的全局地图都是无偏的.在实验中通过实际环境的RGB-D(彩色-深度)数据验证了算法的有效性.

In map merging in multi-robot cooperative SLAM （simultaneous localization and mapping）, global map con- struction maybe fails due to information deficiency caused by limited communication range or communication topology changes of the multi-robot network. To solve the problem, a new dynamic map merging algorithm is proposed based on the consensus of information gain. The proposed algorithm is fully distributed and independent of any specific communication topology. The information gain between the new observed data and the history data of the local map estimated by each robot is calculated and utilized to enable each robot to achieve consentaneous global map simultaneously. The proposed algorithm can asymptotically converge to the true global map under limited communication conditions. Furthermore, the estimated global map of each robot is unbiased in each iteration step. RGB-D （RGB-depth） data collected from real world are used to confirm the efficiency of the proposed algorithm.