Q-learning based on strategic artificial potential field for path planning enabling concealment and cover in ground battlefield environments

Path planning in battlefield environments differs from typical path planning because it may not always involve obstacle avoidance. In fact, obstacles can be utilized to hide from enemies or facilitate troop advancement through destruction and relocation. Therefore, path planning in such environments requires specificity. One widely used method for obstacle avoidance is the artificial potential field (APF) method. This study proposes a strategic artificial potential field (SAPF) algorithm that utilizes obstacles for path planning. The Q-learning algorithm is also being researched for its application in learning methods involving path planning and obstacle avoidance. This study proposes a path planning algorithm utilizing SAPF-based Q-learning that takes into account concealment and cover in various terrains. The proposed Q-SAPF algorithm was compared to Q-learning based on the results of repeated experiments and analysis of variance (ANOVA) and the post hoc Tukey technique to examine the statistical significance and differences between the algorithms, respectively. Q-SAPF outperformed Q-learning by generating shorter path plans with differences in path lengths of 1.96–8.77. It further achieved faster learning times of approximately 14.57–69.29 s and path scores that were higher by approximately 2.73–47.96. Specifically, its success rate was 13.73–54.89% higher than that of Q-learning. The achieved outcome demonstrated the superior learning ability of the Q-SAPF. The study results confirm the feasibility of path planning that incorporates concealment and cover by utilizing obstacles. The findings of this study can contribute to path planning in various situations beyond wartime environments.