IoT-Enabled Real-Time UAV Path Planning for Dynamic Disaster Response

Autonomous UAVs are vital in post-disaster search and rescue missions, where rapid and intelligent path planning is critical. However, traditional algorithms and standard reinforcement learning (RL) approaches struggle with dynamic hazards and often violate UAV kinematic constraints. We propose PPO + KinOpt, a hybrid framework that combines proximal policy optimization with a curvature-aware optimization layer to enable real-time, physically feasible path planning in dynamic environments. Tested in an IoT-informed disaster simulation environment, our method achieves a compact 34.5 m trajectory, 0.903 path efficiency, and maintains curvature within UAV maneuverability limits, significantly outperforming both classical and RL-only baselines. The approach ensures kinematic feasibility, adaptability, and real-time responsiveness, making it well-suited for IoT-driven aerial disaster response.