A Distributed Task Allocation Algorithm for UAV Swarms in Dynamic Disaster Perception

To address the distributed task allocation challenge for UAV swarms undertaking closed-loop search-and-extinguish missions in dynamic forest fire environments, this paper presents a situation-aware distributed algorithm for search-and-suppression task allocation (DA-SSTA) based on the consensus alternating direction method of multipliers (ADMM). DASSTA achieves decision consistency under non-fully connected topologies, multi-objective cost-coupling conflicts, and real-time dynamic reallocation requirements. A normalized multi-objective framework integrating navigation and fire-extinguishing costs is developed, with a Logistic-based adaptive weight mechanism to adjust mission priorities dynamically. The original mixed-integer problem is converted to convex quadratic programming via variable relaxation and constraint affinization, and the non-negative slack variables handle path-fire task coupling constraints to satisfy the Slater condition for strong duality. A multiparameter adaptive penalty selection (MpAPS) mechanism is proposed to update penalties via spectral radius minimization dynamically, and a damping factor is introduced to suppress oscillations, thereby boosting convergence speed and numerical stability. Simulations show that DA-SSTA achieves markedly higher UAV survival and mission completion rates with control-lable growth in navigation cost than four representative baselines, including consensus ADMM and the multi-agent proximal policy optimization algorithm (MAPPO). As such, it performs robustly in scenarios with a high density of fire points. MpAPS cuts algorithm iterations by 60%, while mitigating penalty-parameter oscillations. The proposed method offers a robust, scalable solution for distributed task allocation for resource-constrained UAV swarms in dynamic disaster scenarios, contributing to autonomous collaborations like search and rescue.