Deep Reinforcement Learning-Based Social Welfare Maximization for Collaborative Edge Computing

Collaborative Edge Computing (CEC) is an effective method that improves the performance of Mobile Edge Computing (MEC) systems by offloading computation tasks from busy edge servers (ESs) to idle ones. However, ESs usually belong to different MEC service providers so they have no incentive to help others. To motivate cooperation among them, this paper proposes a cooperative mechanism where idle ESs can earn extra profits by sharing their spare computational resources. To achieve the optimal resource allocation, we formulate the social welfare maximization problem as a Markov Decision Process (MDP) and decompose it into two stages involving the allocation and execution of offloaded tasks. The first stage is solved by extending the well-known Deep Deterministic Policy Gradient (DDPG) algorithm. For the second stage, we first show that we only need to decide the processing order of tasks and the utilized computational resources. After that, we propose a dynamic programming and a Deep Reinforcement Learning (DRL)-based algorithm to solve the two types of decisions, respectively. Numerical results indicate that our algorithm significantly improves social welfare under various situations.