Litcius/Paper detail

Replanning-Oriented Framework for Efficient Real-Time Decision-Making in Multi-UAV Systems

Xingshuo Hai, Longyan Tan, Qiang Feng, Haibin Duan, Changyun Wen

2025IEEE Transactions on Industrial Informatics13 citationsDOI

Abstract

Efficient real-time decision-making for long-term multiple unmanned aerial vehicles (multi-UAV) missions in geo-distributed environments requires an integrated approach to manage dynamic task demands. We propose a hierarchical dual-layer decision-making framework for multi-UAV mission replanning. The upper layer optimizes multi-UAV deployment using the density-constrained K-medoids clustering and simulated annealing algorithm, achieving globally optimal solutions. The lower layer addresses task assignment via the goal-oriented belief space multiagent reinforcement learning algorithm, which leverages updated belief distributions to mitigate sparse reward and enhance training efficiency. Coordination between the two layers ensures comprehensive coverage of predefined demands while adapting to dynamic events. The effectiveness of the proposed methods is validated through a real-world case study using the 911 call dataset from city emergency services.

Topics & Concepts

Computer scienceReal-time computingEmbedded systemRobotic Path Planning AlgorithmsSimulation Techniques and ApplicationsFault Detection and Control Systems