Indoor Mobile Robot Navigation System Study Based On Depth Vision

With the development of computer vision technology,Visual Simultaneous Localization and Mapping(V-SLAM)and autonomous navigation technology have been becoming the focus of current robot research.There are important theoretical significance and practical application value in the research of indoor mobile robot navigation system based on depth vision.In this paper,the related technologies of indoor mobile robot navigation are studied.After some contrast analysis,depth camera is selected as the visual sensor.And the design of the indoor mobile robot navigation system based on depth vision is completed on the Robot Operating System(ROS).In order to solve the problems of large point cloud registration error,low computational efficiency and poor real-time performance in visual SLAM technology,a visual SLAM algorithm based on camera calibration is proposed.The depth camera is calibrated by using a flat chessboard,and multiple point clouds are aligned by external parameters of the depth camera to obtain a multi-view transformation matrix for global point cloud registration.The ORB feature operator,which is less calculation,is used in the front of the visual SLAM system to reduce the amount of calculation and increase the real-time performance of the system.In the loop detection,the key frames are selected by constructing the visual dictionary tree and changing the similarity score calculation method,which can increase the closed-loop recognition rate of the robot and the camera pose estimation accuracy.Moreover,the map optimization is used in the back-end of the optimization,which can reduce the cumulative error and increase the accuracy of the robot pose.The experimental results show that the improved visual SLAM algorithm meets the real-time requirements of the system well,and it improves the accuracy of point cloud matching and map construction.In path planning of robot,a hybrid improved artificial fish is proposed called Admixture Improved Artificial Fish Swarm Algorithm(AIAFSA)in order to solve the problems of the artificial fish swarm algorithm(AFSA),which is easy to fall into local optimum.And the visual field and step parameter values are fixed which lead to low precision of algorithm optimization.The LPA* algorithm is used to plan the global path preliminarily,and each node in the sub-optimal path is used to determine the number of artificial fish.The path planning algorithm for adaptive behavior is proposed byintroducing the weighting factor ?.Therefore,it avoids the local optimum caused by fixing the parameters of artificial fish's visual field and moving step.Thus,the real-time performance of the algorithm and the quality of the optimal solution are further improved,and the attenuation function ? is used to ensure the balance between global and local path planning.The experimental results show that the global optimal path can be planned by the Admixture improved artificial fish swarm algorithm proposed in this paper,which improves the convergence speed and accuracy,and makes the robot navigation path smoother.The construction and implementation of indoor mobile robot navigation system based on depth vision is completed on the laboratory hardware platform.The feasibility and stability of the indoor mobile robot navigation system based on depth vision are proved by experiments in different environments.