AERAS: Adaptive AI-Driven Resource Allocation for Mission-Critical Tactical Edge Computing

Abstract

Mission-critical tactical edge systems require re-source allocation that maintains low latency, energy efficiency and SLA compliance under bursty workloads and heteroge-neous nodes. This paper presents AERAS (Adaptive Attention-Enhanced Reinforcement Allocation System), which combines GRU-based workload forecasting, attention-weighted task–node encoding and PPO-based multi-objective scheduling for closed-loop edge orchestration. The proposed framework jointly opti-mizes latency, energy consumption, throughput and SLA satis-faction through a normalized reward design and is evaluated across high-load ISR, drone-swarm coordination and cyber-defence alert scenarios. Simulation results show that AERAS consistently outperforms Round Robin, First Fit and DQN baselines. In the high-load ISR case, it achieves 0.62 s mean latency and 4.8 J/task energy, while also reaching 94.0% and 90.5% SLA satisfaction in the drone-swarm and cyber-defence scenarios, respectively. These results indicate that predictive and attention-guided reinforcement learning is a practical approach for mission-aware tactical edge scheduling. Index Terms—Edge computing; tactical systems; adaptive scheduling; reinforcement learning; workload forecasting; atten-tion mechanism; defence AI; PPO.