Yingze Su

Efficient path planning serves as a fundamental requirement for autonomous unmanned aerial vehicles (UAVs), especially in low-altitude environments subject to significant meteorological perturbations. While traditional methods prioritize single-objective optimization, non-stationary airspace demands a trade-off between latency, energy, and safety margins. We propose a heuristic-assisted multi-objective path planning framework, named Dynamic Meteorology-Physical Asynchronous Advantage Actor-Critic (DMPA3C), as the implemented Artificial Intelligence (AI) method, specifically designed for UAV trajectory optimization in non-stationary airspace. DMPA3C integrates meteorological constraints into the flight-environment model and establishes a meteorology-guided multi-agent weakly supervised learning network, thereby enhancing UAVs' real-time decision-making capability. First, we develop a three-dimensional dynamic meteorological-physical traversability cost grid (3DMPTG) that fuses multiple meteorological metrics with geographic grids. 3DMPTG characterizes meteorological non-stationarity through temporal differences and spatial gradients, thus translating the resulting impacts on UAV power and velocity into spatiotemporally continuous traversability costs. Furthermore, meteorological constraints are integrated into the state space and multi-objective reward function, which promotes environmental awareness and adaptive control. Finally, we employ a hybrid heuristic strategy for action space shaping and policy guidance, functioning as a weak supervision signal to accelerate convergence and reinforce robustness against dynamic perturbations. Comprehensive comparisons with several advanced algorithms highlight the superior optimization performance of DMPA3C. Specifically, the algorithm maintains a success rate of 96.4% under extreme compound disturbances compared to 82.6% and 87.4% for commercial benchmarks. These improvements demonstrate AI's potential for UAV multi-objective path planning, enhancing efficiency and accuracy under complex meteorological conditions.