Artificial Potential Field based Improved JPS in Weighted Maps

Path planning refers to finding a collision free path from the starting state to the target state in an environment with obstacles, according to certain evaluation criteria. The Jump Point Search (JPS) algorithm, an enhancement of the A∗algorithm, has demonstrated substantial gains in reducing search node numbers and accelerating the search process. However, its efficacy is constrained in weighted maps, necessitating the introduction of APFW-JPS in this paper. APFW-JPS refines the heuristic function using the artificial potential field method and incorporates a neural network to adapt heuristic function coefficients to diverse maps. This augmentation aims to enable JPS to maintain search speed in weighted maps while achieving paths with lower costs. For neural network training, a dataset of 2050 randomly generated maps with varying dimensions and weight distributions was employed. Experiments demonstrate that APFW-JPS effectively diminishes the cost of conventional JPS in weighted maps, concurrently upholding an accelerated search pace.