A Two-Layer Framework Integrating Bidirectional A* and Genetic Algorithm for Low-Altitude UAV 3D Path Planning

Aiming at the operational challenges of low-altitude UAV path planning in complex 3D environments—where rapid feasibility, trajectory smoothness, and safety margins must be balanced under practical computational budgets—this paper proposes a two-layer collaborative framework (GA-B-A*) integrating bidirectional improved A with a customized genetic algorithm. The key technical innovations are threefold: (i) a bidirectional A* variant with a composite cost function that encodes terrain-aware safety margins, reducing search space by approximately 50% compared with unidirectional search; (ii) a customized genetic algorithm with domain-specific operators (path-shaping mutation, altitude-correction mutation) that preserve path continuity during evolution; and (iii) a two-stage smoothing-and-verification pipeline that enforces kinematic feasibility (pitch angle ≤ 30°, curvature constraints) on the final trajectory. Across six scenarios with 0/100/300 obstacles, GA-B-A* shortens path length by 20%~30% (Cohen’s d = 1.90, 95% CI: [2.87, 3.97]), reduces maximum turning angle by ~50% (Cohen’s d = 4.52), and lowers the defined path cost by 15%~18% compared with A*, with all improvements statistically significant at p < 0.001. All paths satisfy pitch angle and curvature constraints with 100% compliance.