Dynamic Path Planning using a modification Q-Learning Algorithm for a Mobile Robot

^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

Robot navigation involves a challenging task: path planning for a mobile robot operating in a changing environment. This work presents an enhanced Q-learning based path planning technique. For mobile robots operating in dynamic environments, an algorithm and a few heuristic searching techniques are suggested. Enhanced Q-learning employs a novel exploration approach that blends Boltzmann and ε-greedy exploration. Heuristic searching techniques are also offered in order to constrict the orientation angle variation range and narrow the search space. In the meantime, the robotics literature of the energy field notes that the decrease in orientation angle and path length is significant. A dynamic reward is suggested to help the mobile robot approach the target location in order to expedite the convergence of the Q-learning and shorten the computation time. There are two sections to the experiments: quick and reassured route planning. With quickly path planning, the mobile robot can reach the objective with the best path length, and with secure path planning, it can avoid obstacles. The superior performance of the suggested strategy is quick and reassured 8-connection Q-learning (Q8CQL) was validated by simulations, comparing it to classical Q-learning and other planning methods in terms of time taken and ideal path.