UAV-Assisted Geo-Edge Routing Protocol for Internet of Vehicles Based on Reinforcement Learning Strategy

With the rapid development of internet of vehicles, the role of mobile edge computing is becoming more and more significant. Mobile users can get massive computing and storage resources locally, thus effectively solving the congestion problem of core network. However, how to deliver data in the complex urban environment becomes a challenge. In this paper, we propose a novel assisted geo-edge routing protocol for internet of vehicles based on reinforcement learning strategy using a neural network-based DDQN (Double Deep Q Network) to implement. The routing protocol is divided into two parts, the first part is the aerial part: the UAV swarm evaluates the suitability of the forwarding section based on the relative speed, number and distribution density of vehicles through a fuzzy logic algorithm, and forwards to the requesting vehicles on the ground in real time. The second part is the ground part: a candidate mechanism is proposed so that the vehicles can apply the DDQN algorithm, learn the input state and use an experience pool to increase the utilization of samples, and select the next hop adaptively. Simulation results show the method performs very well in terms of packet delivery rate and end-to-end delay, which can improve network performance and reduce communication overhead.