多机协同的类脑同步定位建图方法

Multi-robot Cooperative Brain-inspired SLAM

低计算量和存储量的优势使得类脑导航模型RatSLAM适用于大型环境的地图构建。为了进一步提高其建图效率,本文提出了一种多机器人协同RatSLAM系统。首先,设计使用了集中式的通信系统用于机器人间信息的交互。其次,提出了一种环境重叠区域检测方法以实现机器人之间的数据关联,从而进行经验节点间相对位姿计算。最后,提出一种改进型图松弛算法,利用机器人之间的位姿关系进行多机器人地图融合,完成全局统一的经验认知地图的实时在线构建。在公开数据集上以及真实物理实验环境中验证了所提方法的有效性。实验结果表明,相比单机器人,多机器人在保持较高地图精度与较少存储量的同时,平均建图效率提升45%。进一步与当前其他地图融合方法进行了比较,本文方法取得了更好的建图结果,验证了所提方法的有效性和准确性。

The advantages of low computing and storage requirements make RatSLAM,a brain-inspired navigation model,well-suitable for constructing large-scale environmental maps.To further improve the mapping efficiency,a multi-robot collaborative RatSLAM system is proposed.Firstly,a centralized multi-robot communication system is designed.Secondly,a method for detecting environment overlapped region is proposed to achieve data association between robots,so as to calculate the relative pose between experience nodes.Finally,the proposed improved graph relaxation algorithm is used to fuse the maps of multiple robots using the pose relationship between robots,to complete the real-time online construction of a globally unified experience cognitive map.The proposed method is verified on both public datasets and in a real-world physical environment to show its effectiveness.The experimental results show that,in comparison to single robot,the average mapping efficiency of multiple robots is increased by 45%,meanwhile maintaining higher map accuracy and requiring less storage.Furthermore,the proposed approach yields a much better mapping results compared with the existing map fusion methods,confirming its effectiveness and accuracy.