Fuzzy Deep Reinforcement Learning based Space-Air-Ground Integration for Electric Power Emergency Communication

With the advancement of unmanned aerial vehicles (UAVs) and artificial intelligence, space-air-ground integration reshapes electric power emergency communication through coordinated resource scheduling. However, existing learning-based resource scheduling has slow convergence, local optimality, and poor quantifiable interpretability. In this paper, we propose a fuzzy deep reinforcement learning-based resource scheduling design, aiming to maximize the weighted sum of inspection and coverage utility minus backhaul delay cost by optimizing UAV path planning, backhaul routing, and flow control. It develops a multi-agent interactive deep fuzzy neuro actor critic mechanism for collaborative 3D path planning and a backpressure-aware approach for multi-hop data backhaul. Simulations in flooding areas show its superiority in real-time situational awareness, on-demand coverage, and fast inspection.