Keyframe Tracking-based Path Planner for Vision-based Autonomous Mobile Robots

Abstract

Recently, visual navigation systems have been actively studied in mobile robot navigation. Such systems create keyframes to correct and track the pose of a mobile robot. However, unlike manual control, the mobile robot under autonomous control is likely to fail to follow the keyframe path accurately due to control errors. To deal with this problem, we propose a novel local path planner called a keyframe tracking-based path planner (KTPP) that helps a robot to track the keyframe path continuously. The KTPP constantly monitors whether or not the robot is on the keyframe path and if not, a local path is generated to guide a robot to return to the desired keyframe path. Various experiments show that the KTPP lead the robot to arrive at the goal point more accurately.