Research And Realization On Visual SLAM Of Indoor Mobile Robot's Navigation System

Autonomous navigation is the core technology of mobile robot,mainly including mobile robot visual Simultaneous Localization and Mapping(SLAM)and path planning.At present,this is the key research in the field of mobile robot,so the indoor mobile robot navigation system based on visual SLAM has important theoretical significance and practical application value.Firstly,the research status of mobile robot navigation,visual SLAM and path planning related technology is analyzed.The system framework of visual SLAM is designed and the visual sensor is selected.Combined with the research of path planning technology,the overall design of mobile robot visual SLAM navigation system in indoor environment is completed.In the study of visual SLAM,for dealing the problems of large error and low efficiency of point cloud registration that improved the point cloud registration strategy is proposed at front-end.The improved algorithm used the Random Sample Consensus(RANSAC)sampling strategy to screen the RGB figure to obtain the inner points and complete the pre-processing.The point cloud initial registration is completed by using the distance threshold between corresponding points based on the consistency of rigid body transformation.In order to ensure a good initial pose that introduced of a dynamic angle factor Iterative Closest Point(ICP)algorithm to realize the point cloud accurate registration under the condition of the original position is well.Then,an improved loopback detection strategy was proposed to deal the problem that inconsistent map construction caused by cumulative drift in visual SLAM system.The strategy improves the similarity score function of the Bag of Words(BOW)and reduces the ambiguity of the loop detection,the accuracy of loop detection is improved.In order to reduce the comparison times of key frames,in the selection of key frames introduced a double weight strategy based on rotation degree and translation degree.Combined with General Graph Optimization(G2O)algorithm to optimize robot trajectory and achieve global consistent visual SLAM system.An Improved Adaptive Artificial Fish Swarm Algorithm(IA-AFSA)is proposed to solve the problem of optimal path planning in indoor environment.An adaptive visual and step size strategy is proposed to balance the global and local search capabilities.The weight evaluation factor is introduced into the foraging,clustering and following behavior of the fish,which can better select the optimal behavior of the fish and effectively avoid the algorithm falling into the local extreme value and premature state.Experiments show that the improved algorithm can effectively improve the convergence speed and accuracy of the algorithm and plan the optimal path.Finally,using the Pioneer3-DX mobile robot to build a test platform for visual navigation system based on depth camera,and the software and hardware is designed for the system.The improved visual SLAM and path planning algorithm is tested and analyzed.The experimental results show that the improved algorithm is effective and the indoor mobile robot's navigation system based on visual SLAM has higher reliability and feasibility.