Uncertainty-Aware MPC Routing for UAV-Assisted VANETs using RNN-GANs

Abstract

The huge Internet of Things (IoT) that 6G envisions needs transmission systems that are smart, self-sufficient, and long-lasting. It’s hard to meet the strict Quality of Service (QoS) needs of future applications when there is a lot of movement and the network conditions are hard to predict, like Vehicular IoT (V-IoT). This study presents an AI-driven, cross-layer control architecture that uses dynamic aerial edge nodes (UAVs) to provide reliable data distribution. A Recurrent Neural Networkbased Generative Adversarial Network (RNN-GAN) is essentially a prediction engine that forecasts the probability distributions of future link quality and network congestion. These observations from different layers are fed into a Model Predictive Control (MPC) agent at the network layer. This agent chooses routes ahead of time while taking risk into account. Preliminary simulation results show that the suggested strategy improves Packet Delivery Ratio (PDR) by up to 16%, lowers end-to-end latency by over 30% with high traffic, and up to 65% when employing aerial relays.