Optimized path planning for underground coal mine filling: an greedy adaptive directional bidirectional rapidly-exploring random tree

Abstract The technology of mining filling is of great significance in improving coal recovery rates, protecting the environment, and conserving land resources. The current efficiency of filling is constrained by single-method approaches. To address this issue, this study develops a path planning model based on underground fill space data, which comprehensively considers fill path length and material volume using a goal programming method, and designs corresponding constraints and adaptive weights. To further optimize search efficiency, an adaptive directional bidirectional rapidly-exploring random tree (AD-BIRRT) algorithm is proposed. This algorithm can intelligently adjust the optimal exploration direction based on current demand and state, significantly enhancing search efficiency and accuracy through the establishment of a dual-tree structure. Additionally, a novel greedy strategy is introduced to resolve path smoothing and redundancy issues. To verify the rationality and effectiveness of the proposed method, comparative tests were conducted in test scenarios, fill scenarios, and on experimental platforms against BIRRT, BIRRT*, Genetic algorithm, and artificial potential field algorithms. The results indicate that the proposed greedy AD-BIRRT algorithm exhibits significant advantages in terms of computation time, path quality, and material accumulation. This algorithm effectively enhances the efficiency and quality of the filling process.