Learning Based Massive Data Offloading in the IoV: Routing Based on Pre-RLGA

Abstract

With the increasing demand for bulk data transmission, offloading techniques to transmit data via mobile physical media is becoming common increasingly. The development of the Internet of Vehicles (IoV) provides new solutions for data offloading, as the IoV brings massive storage and transmission capabilities. However, the impact caused by the mobility of physical media brings a challenge to delay optimization and route selection of data offloading. In this paper, we consider a data transmission network architecture based on the Manhattan mobility model. The vehicle carries data on a fixed route in this scenario, enabling data transmission between geographically distant data centers. In order to reduce the total transmission time and reduce the impact of retransmission, we consider the temporal convolutional network (TCN) model to predict the allocation of the weight of delay. Next, we solve the optimal routing problem using a genetic algorithm based on a reinforcement learning mechanism (RLGA) to pre-allocate resources for offloading requests. The experimental results show that the proposed data offloading method can reduce the load on the cellular network and decrease the data transmission time, average transmission hops, and retransmission times compared with existing methods.