Dual-stage collaborative path planning and task allocation for UAV swarms in complex maritime rescue environments

Abstract

The growing complexity and urgency of maritime rescue missions highlight the limitations of traditional multi-UAV coordination methods under nonlinear constraints and uncertain task failures, calling for a more robust and efficient joint optimization framework. This study proposes a novel dual-stage collaborative optimization framework to tackle the coupled problem of dynamic path planning and fault-tolerant task allocation. In the first stage, an improved dream optimization algorithm (IDOA) incorporates differential evolution-based perturbation and a local memory mechanism to enhance global exploration and path robustness in complex wind-terrain environments. In the second stage, an adaptive fault-tolerant cooperative multi-task assignment problem (AFT-CMTAP) model is developed for heterogeneous UAVs, and a discrete variant of the dream optimization algorithm (DIDOA) is designed to solve this combinatorial problem using genetic-inspired operators. The framework supports online replanning through real-time risk evaluation and adaptive resource scheduling, ensuring mission continuity and system-level reliability under failure conditions. Experimental results demonstrate that the proposed IDOA+DIDOA framework reduces task completion time by up to 37.4 % and decreases path cost by 28.7 % compared to baseline methods. It maintains robust under multiple simultaneous failures, showing strong adaptability and stability in uncertain environments, which supports its broad applicability in cross-regional safety and reliability planning.