Deep Reinforcement Learning-Enhanced Overlapping Coalition Formation with Blockchain-Secured Trust Management for Multi-UAV Networks
Iftikhar Rasheed, Hala Mostafa, Saad Alahmari, Saad Nasser Altamimi, Ahmed Thair Shakir
Islamia University of Bahawalpur Princess Nourah bint Abdulrahman University Northern Border University Imam Mohammad ibn Saud Islamic University
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
This paper presents a novel deep reinforcement learning-enhanced overlapping coalition formation game (DRL-OCFG) framework for multi-UAV networks that jointly optimizes task allocation, power management, and trust relationships. Unlike existing approaches that rely on static preference orders and centralized coordination, our framework introduces three key innovations: (1) a multi-agent deep deterministic policy gradient (MADDPG) approach with graph neural network integration that achieves dynamic coalition adaptation in real-time, reducing convergence time by 52% compared to traditional game-theoretic methods; (2) a blockchain-based trust management system using a lightweight consensus protocol with hierarchical sharding that ensures coalition integrity with only 26ms latency overhead while preventing Byzantine attacks and scaling gracefully beyond 200 UAVs; and (3) an adaptive energy harvesting and power optimization mechanism with time-varying environmental modeling that extends operational duration by 145% through joint trajectory and transmission power control, including night-time and adverse weather operation modes. We formulate the problem as a constrained Markov decision process incorporating dynamic task arrivals and propose a tiered computation architecture that enables deployment across heterogeneous UAV platforms ranging from low-cost micro-UAVs to high-capability platforms. Extensive simulations across diverse environments—including dense urban, extreme weather, and signal-denied scenarios—demonstrate that the proposed DRL-OCFG framework achieves 46% improvement in task completion rates, 39% reduction in energy consumption, and maintains stable performance with up to 500 UAVs. Real-world experiments using a heterogeneous swarm of 12 UAVs with extreme capability gaps validate the practical feasibility, showing successful deployment in disaster response scenarios with 96% mission success rate under dynamic environmental conditions and dynamic task arrivals.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
工程UAV Applications and Optimization
Distributed Control Multi-Agent Systems · Reinforcement Learning in Robotics