Path Planning Improvement Using a Modified Q-learning Algorithm Based on Artificial Potential Field
International journal of intelligent engineering and systems · 2024
Over recent years, the demand for mobile automated navigation has grown significantly.Path planning has emerged as an important and exciting field of research in many disciplines, intending to create shorter and smoother paths through the use of several types of algorithms.This work presents a comparative analysis between the artificial potential field (APF) method and the hybrid approach that combines Q-learning with APF (QL-APF).Subsequently, these methodologies are juxtaposed with a proposed approach, modified Q-learning with APF (MQL-APF), which introduces modifications to QL-APF by incorporating a dynamic reward function with a static reward function.The proposed approach has been empirically proven effective, and it is capable of generating safe and efficient paths even in complex environments.The MQL-APF approach achieved improvements in terms of path length of approximately 67.25% when compared with the APF method.Furthermore, the average enhancement percentage is approximately 14.68% compared to the QL-APF method.Additionally, the MQL-APF method achieved an improvement of 50.15%, 7.21%, and 24.63% for QL, MQL, and QL-APF, respectively.