AI-Driven UAV-Assisted Edge Computing for Rapid Response in Emergency Wireless Networks

Abstract

In this study, the researcher intends to test the design of an AI-powered UAV-based edge computing application that will facilitate faster and reliable wireless communication in emergency settings, including natural catastrophes or infrastructure failure. The terrestrial networks in most cases do not work under such conditions and an aerial solution that is intelligent and flexible is therefore necessary to work in real time processing high data and restoration of the networks. The system proposed introduces autonomous swarms of UAVs capable of edge computing nodes and deep reinforcement learning (DRL) to optimize the trajectories of the UAVs, as well as distribution of both computing tasks and routing communication dynamically. Each UAV is used as a mobile edge node and all of them create a self-organizing aerial network which is flexible about the changes in terms of user demand, topology, and energy limitations. The DRL model was Proximal Policy Optimization (PPO) based, and simulations were done in a 4-km by 4-km disaster area. Findings show that decision latency is 74 percent shorter, network throughput is 61 percent higher, and coverage loss is 5.2 percent instead of static base stations and standard mesh networks. This UAV AI-based design can provide a scalable and robust low latency, high reliability communication service within the category where the infrastructure does not exist and can solve the gap between the ground users and the computational services. Future development will entail satellite connectivity using the model as well as multi-
modal sensor fusion extension of the model.