Research On Active Collaborative Detection Technology Of Multi-Robot Based On SLAM

In recent years,the exploration of complex unknown environment,such as unknown planet surface,dense jungle,battlefield environment and post-disaster surface environment,has become one of the research hotspots of multi-robot systems.Multi-robot system has the characteristics of high efficiency and strong robustness.On the basis of cooperation,it is more suitable for the detection of complex unknown environment.Aiming at the Multi-robot active cooperative detection task in complex and unknown environment,this paper studies the system framework of multi-robot active cooperative detection environment,the multi-robot map fusion algorithm based on RGB-D SLAM,the multi-robot cooperative positioning algorithm integrating relative observation information,the multi-robot map fusion experimental platform and other content.First,this paper design the multi-robot collaborative detection environment system actively,including overall control planning subsystem,local control planning subsystem and robot subsystem,can realize the complicated unknown environment map construction and high precision positioning of the multi-robot system function.Secondly,a map fusion algorithm is proposed to provide initial transformation matrix of ICP by judging the map overlapping area against the global map module of multi-robot active cooperative detection system,which solves problem of complex initial state setting of map fusion algorithm.On the basis of RGB-D SLAM mapping,the overlapping of maps is found by the feature point matching and contrapolar constraint,and the initial transformation matrix between maps is calculated.After ICP algorithm iteration,the accurate transformation matrix between maps is obtained.The simulation results show that the average accuracy of the multi-robot map fusion algorithm can reach 6.4%.Thirdly,a multi-robot collaborative positioning algorithm based on relative observation information fusion is proposed against the multi view collaborative positioning module.On the basis of filter generation algorithm,the relative observation information is fused with the system state and covariance error correction link to reduce the Markov hypothesis error caused by linear positioning and the cooperative positioning accuracy of multi-robot system is improved.The simulation environment is built to verify that the error value of relative observation information is within 10 cm,and the positioning accuracy of fused relative observation information is higher than that of non-fused relative observation information.Finally,for the multi-robot map fusion algorithm,an experimental environment composed of mobile robot,depth camera and central computer is built to verify the algorithm.The result show that the multi-robot map fusion algorithm can judge whether there are overlapping areas between local maps in indoor environment,and can effectively fuse local maps on the basis of having overlapping areas.The average relative error of map fusion is3.6%.