A Scalable Tree Based Path Planning for a Service Robot

Abstract

Path planning plays a vital role in a mobile robot navigation system. It essentially generates a shortest traversable path between two given points. There are many path planning algorithms proposed by researchers all over the world, however, there are very little work focussing on path planning for a service environment. The general assumption are either the environment is fully known or unknown. Both the cases will not be suitable for a service environment. A
fully known environment will restrict further expansion in terms of number of navigation points and unknown environment would give an inefficient path. Unlike other environments, service environments have certain factors to be considered, like user friendliness, repeatability, scalability and portability which are very essential for a service robot. In this paper a simple, efficient, robust and environment independent path planning algorithm for an indoor mobile service
robot is presented. Initially the robot is trained to navigate to all the possible destinations sequentially with minimal user interface which will ensure the robot knows partial paths in the environment. With the trained data the path planning algorithm maps all the logical paths between all the destinations, which helps in autonomous navigation. The algorithm is implemented and tested using 2D simulator Player/Stage. The proposed system is tested with two different
service environment layouts and proved to have features like scalability, trainability, accuracy and repeatability. The algorithm is compared with various classical path planning algorithms and the results shows that proposed path planning algorithm is in par with the other algorithms in terms of accuracy and efficient path generation.