PREDICTIVE INTELLIGENCE DRIVEN GRAPH NEURAL REINFORCEMENT FRAMEWORK FOR RESILIENT HIGH MOBILITY MANET TOPOLOGIES

Mobile Ad Hoc Networks (MANETs) are highly dynamic wireless communication systems characterized by decentralized architectures, continuous node mobility, and frequent topology variations. Traditional reactive routing protocols often fail to maintain reliable communication under high-mobility conditions due to delayed responses to link failures and inefficient energy management. This study proposes an intelligent predictive framework integrating Graph Neural Networks (GNN), Gated Recurrent Units (GRU), and Multi-Agent Deep Reinforcement Learning (MADRL) to enhance resilience, routing efficiency, and energy optimization in MANET environments. The research systematically evaluates the effectiveness of predictive intelligence in improving link stability, packet delivery, latency reduction, and network adaptability. Experimental analysis was conducted using statistical evaluation, correlation analysis, and model fit assessment techniques. The findings reveal that AI-driven predictive topology control significantly improves network resilience and communication reliability compared with traditional reactive routing approaches. The proposed framework also demonstrates strong scalability for real-time applications including UAV swarms, disaster recovery communication, and military networking systems. The study highlights the importance of integrating predictive AI mechanisms into next-generation decentralized wireless communication architectures.