Learning-Based Cooperative Cache Strategy Integrating Content Recommendation in IoV

Abstract

As the proliferation of in-vehicle information services continues to grow, the size of cached content expands, consequently extending the caching duration for vehicle users. Internet of Vehicles (IoV) technology can provide a multitude number of communication services between vehicles and Roadside Units (RSUs), making vehicle cooperative caching a viable solution. Nevertheless, for new energy vehicles such as electric ones, how to reduce the energy consumption of the vehicle network system remains a formidable challenge. Additionally, the task delay of vehicle cooperative caching can also be further reduced. In this paper, a clustering caching model is proposed by considering constraints for tolerable delay of tasks and the residence time of vehicles, and the kinetic energy recovery scheme of vehicles and the Energy Harvesting (EH) mechanism of RSUs are designed to reduce the system energy consumption. Terahertz (THz) communication is employed in vehicle to vehicle (V2V) to accelerate the completion of tasks. Subsequently, the energy consumption optimization problem for the vehicle cooperative caching system is constructed. Furthermore, the continuity and high-dimensionality properties of the decision variables for the optimization problem are analyzed, and we propose the joint Download Size and Bandwidth Allocation (DSBA) algorithm to explore an approximate optimal solution for task decomposition and resource allocation in the high-dynamic environment. We also propose an algorithm named Vehicle Cache Content Recommendation (VCCR) to avoid the additional delay and the waste of energy caused by the vehicle user cache miss-hit problem. The simulation results demonstrate that our proposed algorithms have significant advantages in saving task delay and system energy consumption compared to other solutions.