Modifiedment the Performance of Q-learning Algorithm Based on Parameters Setting for Optimal Path Planning

^1,3,4 Electrical Engineering Department, University of Technology - Iraq
² Computer Science Department, University of Technology - Iraq

^* Corresponding author: Eee.21.14@grad.uotechnology.edu.iq

Abstract

In engineering, the use of mobile robots to teach automatic control is becoming more common because of the interesting experiments that can be conducted with them. In this paper, a mobile robot that applies reinforcement learning in different scenarios is shown, to get rewards, the agent learns by acting in the environment. creating a balance between new information and our current understanding of the environment. In this way, the algorithm can be divided into two stages: the learning stage and the operational stage. In the first phase, the robot learns how to go from where it is to a known destination, it builds a learning matrix that is subsequently utilized during the operational stage using the rewards and environment data. In this paper, the algorithm was studied in terms of rapid learning for the mobile robot and reducing the process of repetition in learning by specifying the values of alpha (α) and gamma (γ) in a way that is appropriate for preserving the variance and differentiation between them. To evaluate the robot’s adaptability to various dynamic situations, several simulated test scenarios were executed. In the testing situations, several target motion kinds and numbers of obstacles with various dynamicity patterns were used. The test scenarios illustrated the robot’s adaptability to various settings.