Visual navigation for mobile robot with Kinect camera in dynamic environment

Abstract

To solve the problem of the visual navigation for mobile robot in dynamic environment, a visual navigation system for mobile robot with Kinect camera is designed. Firstly, the improved RBPF (RAO Blackwellized particle filters) algorithm is used to build the 2D grid map of the indoor environment and on the basis of this map, operation of inflating obstacles is applied to build the global grid map considering the actual size of the robot. Secondly, cost map using motion primitives is built while using this cost map, anytime Repairing A*(ARA*) global path planning algorithm, which has property of anytime algorithm, is combined with Dynamic Window Approach (DWA ) local path planning algorithm to plan a smooth path from start point to target point and generate the optimal control input for robot motion. Finally, the adaptive monte carlo localization method (KLD-Sampling) is used to locate the robot and then a visual navigation system for mobile robot with Kinect camera is designed. The indoor mobile robot navigation experiment results show that the designed robot navigation system can plan a smooth path which is in accordance with robot kinematics and autonomously avoid the static and moving obstacles in the environment.