An improved dung beetle optimizer for three-dimensional UAV path planning
Abstract
Three-dimensional UAV path planning in complex obstacle environments is a high-dimensional, nonlinear, and strongly constrained optimization problem, in which flight safety, path length, maneuverability, altitude variation, and energy consumption must be considered simultaneously. To address this problem, this paper first establishes a continuous pathplanning optimization model that integrates path length cost, obstacle threat cost, altitude cost, maneuvering cost, and energy consumption cost, thereby providing a comprehensive evaluation framework for UAV path quality. Based on this model, an improved Dung Beetle Optimizer, named OADBO, is proposed by incorporating low-discrepancy sampling, opposition-based learning, obstacle-aware initialization, multi-leader guidance, historical best memory, velocity-assisted updates, diversity maintenance, stagnation recovery, and local path refinement, so as to improve initial population quality, enhance convergence stability, and reduce the risk of premature convergence. Comparative experiments are conducted in three representative scenarios, namely Corridor, Cluster, and Barrier, against DE, PSO, HHO, and the original DBO. The experimental results show that OADBO achieves the best cross-scenario average performance, with an average cost of 195.50, reducing the cost by approximately 81.6% compared with the original DBO, and demonstrating better solution quality, robustness, and environmental adaptability for three-dimensional UAV path planning.