A multi-stage hybrid path planning framework for UAV nodes in emergency delay-tolerant networks

To restore urban emergency communication networks disrupted by infrastructure damage, Unmanned Aerial Vehicles (UAVs) can be leveraged to construct a Delay-Tolerant Network (DTN), termed U-DTN. The performance of UAV-assisted DTNs strongly depends on path planning, which requires balancing global path optimality and real-time flight safety. This paper proposes a multi-stage hybrid framework for U-DTN node path planning. In the offline phase, a hybrid swarm intelligence algorithm integrating Cuckoo Search (CS) and Grey Wolf Optimizer (GWO), denoted as CSGWO, generates an initial global path. The path is refined by curvature-based pruning and cubic B-spline smoothing to produce a Global Reference Trajectory (GRT) that balances path-length optimality and dynamic feasibility. In the online phase, a Guided Artificial Potential Field (G-APF) method enables fast local obstacle avoidance while maintaining guidance from the GRT. Simulation results show that the proposed framework outperforms baseline methods. It maintains a 100% path generation success rate in offline planning, generates shorter and smoother trajectories, and achieves a 100% unknown-obstacle avoidance success rate online with lower replanning time and smaller path increment cost. These gains enhance U-DTN connectivity and robustness.