Multi-Task Driven User Association and Resource Allocation in In-vehicle Networks

With the rapid development of intelligent vehicles, heterogeneous in-vehicle networks (HetIVNets) applying heterogeneous access technologies in terms of cellular and in-vehicle WiFi, are widely employed to provide stable and ubiquitous network environments for intelligent vehicles and their passengers. Most existing studies on the optimization of heterogeneous networks (HetNets) focus on user association, channel and power allocation. Additionally, these studies typically employ a single task metric to characterize the quality of service (QoS) requirements of the devices. In order to achieve green and energy-efficient intelligent vehicles, our work considers a HetIVNet composed of WiFi and cellular networks, and further optimizes the bandwidth allocation and energy consumption of WiFi access point (AP). Furthermore, to achieve more accurate resource allocation for different tasks in HetIVNet, we establish the QoS requirement model of various tasks for in-vehicle devices. Since the proposed optimization problem is non-convex and NP-hard, we formulate the objectives and constraints of user association and resource allocation (UARA) in HetIVNets as a markov decision process (MDP) and propose a proximal policy optimization (PPO) algorithm for in-vehicle intelligent resource allocation to uniformly schedule network resources. Simulation results validate that the proposed algorithm can achieve high task success rates under low-energy consumption conditions for WiFi AP. We also compare our algorithm with the state-of-art baselines to highlight its efficiency and stability.