RIS-UAV Joint Emergency Rescue Communication Network Performance Based on an Improved TD3 Scheme
TL;DR
A STAR-RIS-assisted UAV emergency rescue network is developed and the method jointly controls UAV deployment, RIS phase shifts, and communication-sensing power allocation under time-varying obstruction, demonstrating robust integrated communication and sensing performance while maintaining competitive energy efficiency under severe dynamic obstruction.
Abstract
Post-disaster environments involve severe blockage, strong scattering, deep shadowing, and heavy-tailed fading, making reliable emergency communication and life-sign sensing difficult. This study develops a STAR-RIS-assisted UAV emergency rescue network and models the control-center-to-UAV link using power-law path loss and Rician fading, while the UAV-to-user and UAV-to-target links are characterized by the Fisher–Snedecor F distribution to capture debris obstruction and multipath effects. To jointly optimize communication throughput and life-detection accuracy, an improved GD-TD3 algorithm is proposed by integrating a GRU-based temporal memory module, Softmax value smoothing, and Gaussian action exploration. The method jointly controls UAV deployment, RIS phase shifts, and communication-sensing power allocation under time-varying obstruction. Simulation results show that activating only 50% activation ratio preserves approximately 81% of the sensing performance achieved by the Full-array STAR-RIS benchmark. In dynamic high-obstruction scenarios, GD-TD3 achieves about 72% communication coverage, a 68% sensing detection rate, and an 87% system detection rate while maintaining 3.4–3.8 Gbps throughput and 0.5594 Mbps/J energy efficiency. These results demonstrate robust integrated communication and sensing performance while maintaining competitive energy efficiency under severe dynamic obstruction.