An Adaptive Sliding-Window Ant Colony Optimization Approach for Multi-UAV Delivery Routing with Safety and Lateness Objectives
TL;DR
Simulation experiments demonstrate that SW-MSACO achieves improved Pareto solution quality and search stability compared with existing heuristic optimization approaches, particularly under large-scale and high-load scenarios, confirming the effectiveness of the proposed framework for complex UAV logistics optimization.
Abstract
Urban unmanned aerial vehicle (UAV) logistics systems require routing algorithms that can balance delivery efficiency and operational safety under complex physical and environmental constraints. However, existing multi-objective routing methods often show limited adaptability when the search landscape changes during optimization. In particular, conventional ant colony optimization (ACO) algorithms usually rely on fixed pheromone update rules, which may lead to premature convergence, uneven Pareto front distributions, and pheromone-induced search stagnation. To address these limitations, this paper proposes a Sliding-Window Adaptive Multi-Strategy Ant Colony Optimization (SW-MSACO) framework for complex urban UAV delivery routing. First, a Complex Multi-objective Urban Routing Problem (CMURP) is formulated by jointly considering cumulative delivery lateness and safety risk under payload and nonlinear battery energy constraints. Second, a multi-strategy pheromone evolution pool is designed to provide complementary search behaviors for convergence acceleration, safety-efficiency balancing, and Pareto front diversity preservation. Third, a Sliding-Window Adaptive Strategy Selection (SW-ASS) mechanism is introduced to adjust strategy selection probabilities according to recent search performance rather than long-term cumulative rewards. Comprehensive simulation experiments under different map scales and task-load conditions demonstrate that SW-MSACO achieves improved Pareto solution quality and search stability compared with existing heuristic optimization approaches, particularly under large-scale and high-load scenarios, confirming the effectiveness of the proposed framework for complex UAV logistics optimization.