A two-stage approach for multi-UAV mission planning in vessel logistics services

This paper proposes a two-stage framework for multi-UAV mission planning in logistics for dynamically moving waterborne vessels. In the first stage, focused on offline spatiotemporal mission planning, the Dynamic Time Warping (DTW) and Partitioning Around Medoids (PAM) algorithms are integrated to extract vessel trajectory features and characterize the spatiotemporal patterns of demand points. Subsequently, a multi-objective optimization model is constructed, which simultaneously minimizes flight distance and energy expenditure, incorporating operational constraints such as UAV performance limitations and mission requirements. A Tent-chaotic Nonlinear Decreasing Crossover-Mutation Gray Wolf Optimizer (TNDC-GWO) algorithm is designed to enhance computational efficiency and convergence performance. Empirical results demonstrate that the proposed TNDC-GWO algorithm yields an 18.1% reduction in total path length and a 15.7% decrease in energy consumption compared to the conventional GWO. In the second stage, designed for online dynamic collision avoidance, an Improved Deep Deterministic Policy Gradient (IDDPG) algorithm, which incorporates hybrid exploration strategies and dual experience replay mechanisms, is proposed to mitigate collision risks from non-cooperative intruder UAVs. Through interactive training and multi-scenario testing, the proposed framework exhibits superior convergence behavior, adaptive path-replanning capabilities, and enhanced generalization performance. The developed methodology not only enhances the operational effectiveness of vessel-based logistics systems through coordinated spatiotemporal planning and real-time collision avoidance, but also establishes an extensible paradigm for multi-UAV cooperative navigation in dynamic environments, highlighting its significant engineering applicability.