Learning Sensor Based Mobile Robot Simultaneous Path Planning and Map Building

Abstract

In this paper, we address the problem of an autonomous mobile robot path planning in an unknown indoor environment. The improved parti-game variable resolution reinforcement Learning approach is applied for planning an obstacle free path from a starting position to a known goal region, and simultaneously build a map of straight line segment geometric primitives based on the application of the Hough transform from the actual and noisy sonar data. The built map is then integrated with the improved parti-game world model, allowing the system to make a more efficient use of collected sensor information. Then an overall improved new method for goal-oriented navigation is presented. It is assumed that the robot knows its own current world location obtained through the accumulation of encoder information, and the robot is able to perform sensor based obstacle detection and motions. Experimental results with a real Pioneer 2 mobile robot demonstrates the effectiveness of the discussed methods.