An Energy-efficient Task Offloading Model based on Trust Mechanism and Multi-agent Reinforcement Learning

Abstract

A task offloading model based on deep reinforcement learning and user experience degree is proposed. Firstly, after users generate blockchain tasks, Proof of Work (PoW) consensus mechanism is introduced to pack the transaction information into blocks to ensure the real reliability of the interaction information in the system. Then, the user experience degree is defined by introducing the total system consumption delay and user gain, and the goal of optimal user experience degree is constructed. Furthermore, the deep reinforcement learning algorithm is used to optimize the offloading model, and the deep reinforcement learning model is constructed by taking the size of the transaction data and the difficulty of the PoW consensus process as the network state, the task offloading and resource allocation relationship between the user and the edge server as the network action, and the user experience as the network reward. Finally, the gradient descent method and back propagation algorithm are used to train the depth network until convergence, and the optimal offloading and resource allocation decision is obtained, and the superiority of the offloading model and algorithm proposed in this thesis is verified by simulation.