Research On Multi-robot Joint Mapping Technology Based On Pose Graphs Fusion

In the future,under the background of a large number of military ground mobile robots will be equipped in the army,the problem of multi-robot cooperative combat becomes more and more important.The relative and absolute location of multi-robot is one of the difficulties faced by multi-robot systems.The joint mapping and location based on multi-robot coordination is an effective method to solve the relative and absolute location of multi-robot.In this paper,the loop-closure detection of multi-robot joint mapping technology and global optimization are studied.The following research results have been achieved.Firstly,the loop-closure detection algorithm of multi-robot joint mapping is studied.The many-to-many matching method based on correlation scan matching has the advantage of computational efficiency,and the closed-loop detection method based on particle filter has the advantage of accuracy.In this paper,the two methods are combined to solve the two key problems of unknown initial relative pose and loop-closure detection in multi-robot mapping.The accuracy rate and detection efficiency of loop-closure detection are organically unified.It provides the basis for global optimization of multi-robot mapping.Secondly,the global optimization algorithm for multi-robot joint mapping is studied.By introducing anchor points and using the factor graph method to establish a joint model of multi-robot pose graphs fusion,the global optimization algorithm based on this model is studied.When the robots meet each other,the robot's pose nodes are optimized by using global optimization algorithm which solves the robot's multiple encounter problems.Global optimization avoids the error accumulation caused by the rigid body transform alone.At the same time,this method solves three problems of poses and map updating,uncertainty of relative pose and computational complexity.Finally,an experimental platform for multi-robot mapping is built,and the multi-robot mapping experiment is carried out by using the above method.A global and consistent map construction in large-scale structured environment is realized.The effectiveness of the multi-robot joint mapping method based on pose graphs fusion is verified.