Resource Allocation Strategy Based on Improved Auction Algorithm in Mobile Edge Computing Environment

Mobile edge computing has emerged as a new paradigm in which compute and storage services from remote cloud data centers are moved to edge cloud servers at the edge of the network. Compared to traditional cloud data centers, edge cloud servers (ECSs) are geographically closer to mobile users and so reduce their communication latency. In this paper, we use auction theory to investigate the allocation of virtual machine (VM) resources in edge cloud servers to mobile users to maximize the profitability of edge servers while satisfying task processing latency and energy consumption constraints of mobile devices. First, we consider mobile users and ECSs as buyers and sellers of VM resource auctions, respectively. Then, we propose a deep learning-based optimal auction, specifically, we build a multilayer neural network architecture based on the parsing solution of the optimal auction. The neural networks first perform monotonic transformation of buyers' bids, and then they compute allocation and conditional payment rules for buyers. We use the buyers' valuations as training data to tune the parameters of the neural network to optimize the loss function, which is the expected negative return of the mobile edge computing service provider. The auction proposed in this paper will not only maximize the profitability of the edge server by satisfying the task processing latency and the energy consumption limit of the mobile device. It also ensures that mobile devices do not unfairly affect the auction results.