Multi-UAV Path Planning Using Improved Artificial Hummingbird Algorithm Based on Differential Evolution and Gradient Descent

This paper proposes a multi-UAV path planning method based on an improved Hummingbird Algorithm (AHA). To address the original AHA’s tendency to become trapped in local optima and its slow convergence, we introduce a novel hybrid algorithm, DGAHA, by integrating Differential Evolution (DE) and Gradient Descent (GD) strategies. The DE strategy enhances population diversity and global search capability, while the GD strategy improves local search efficiency and solution accuracy. Experimental evaluations on the CEC2005 test set (18 functions) and CEC2022 benchmark functions indicate that DGAHA achieves an average solution accuracy improvement of 20%-35% over other algorithms. Applied to the multi-UAV path planning task, the proposed DGAHA method effectively generates feasible and smooth flight paths, reducing path length by 28.5%-44.4% and improving trajectory smoothness by 30.2%-46.3%, with smaller standard deviations and faster convergence speeds, demonstrating significant potential for practical applications.