A Meta-Learning Based Resource Allocation Algorithm in Vehicular Edge Computing Networks

Abstract

With the rapid development of intelligent transportation systems, in-vehicle edge computing networks serve as key infrastructures to provide strong support for vehicle communication and coordination. However, the traditional resource allocation methods face many challenges due to the changing network topology, the uncertainty of data load, and the spatio-temporal relationship between traveling vehicles. Therefore, in this paper, a meta-learning based resource allocation algorithm is designed to overcome these problems. By introducing meta-learning, the aim is to enable the system to quickly adapt to different scenarios and requirements so as to improve the resource utilization. The algorithm not only fully considers the network performance index, but also comprehensively solves the real-time and reliability requirements to achieve comprehensive optimization. The resource allocation algorithm based on meta-learning designed in this paper simulates different car networking tasks during the training process, so that the network learns a more robust resource allocation strategy, which not only fully takes into account the network performance index, but also comprehensively solves the real-time and reliability requirements and achieves comprehensive optimization. Through a large number of simulation experiments and comparisons in Telematics, the algorithm proposed in this paper outperforms the current state-of-the-art algorithms by 20% in terms of request acceptance rate, and leads by 15% in terms of overall gain. The experimental results show that the algorithm in this paper outperforms the current state-of-the-art algorithms in solving the resource allocation problem for car networking applications in edge computing networks.