Abstract: Aiming at  the problem of minimizing latency and energy consumption in service migration in dynamic Internet of Vehicles, we proposed a service migration strategy based on graph neural networks. Firstly, we established a system model to model the service migration problem as a multi-objective optimization problem that minimized latency and energy consumption. Secondly, we  transformed the problem into a Markov decision process, and defined the states, actions, and reward functions. Finally, a graph convolutional neural network was used  to extract features from edge network topology and node information, and combining  task information to propose a deep reinforcement learning task migration decision algorithm based on graph convolutional neural networks to make service migration decisions. Simulation experiment results show that the proposed algorithm outperforms other baseline algorithms in reducing average task latency and energy consumption, and can provide an effective solution for efficient, low consumption, and stable service migration in the environment of Internet of Vehicles.

Key words: mobile edge computing, service migration, graph neural network, Internet of Vehicles