Comparison of SARSA algorithm and Temporal Difference Learning Algorithm for Robotic Path Planning for Static Obstacles

Abstract

For any mobile device, the ability to navigate by its own in its environment is a very important requirement. Avoiding dangerous situations such as accidents and conditions that are not safe comes primarily and if the robot is designed to find certain places according to its application then the robot should find those places. Path planning is an important requirement for such mobile robots. For doing this the robots needs to be aware of their environment. This is possible only if robots are provided with information about the environment. In this paper, the different obstacle occurring cases are studied and simulated for SARSA and temporal difference algorithm for a 6×6 grid in MATLAB and it is found that the success rate of SARSA is much better than that of TD. It is proposed to implement the SARSA algorithm in iRobot with a mobile interface for interaction with the user.