Deep reinforcement learning for the computation offloading in MIMO-based Edge Computing

Multi-access Edge Computing (MEC) has recently emerged as a potential technology to serve the needs of mobile devices (MDs) in 5G and 6G cellular networks. By offloading tasks high-performance servers installed at edge wireless networks, resource-limited MDs can cope with proliferation recent computationally-intensive applications. In this paper, we study computation problem massive multiple-input multiple-output (MIMO)-based MEC system where base stations are equipped large number antennas. Our objective is minimize power consumption delay under stochastic environment. To end, introduce new formulation Markov Decision Process (MDP) propose two Deep Reinforcement Learning (DRL) algorithms learn optimal policy without any prior knowledge environment dynamics. First, Q-Network (DQN)-based algorithm solve curse state space explosion defined. Then, more general Proximal Policy Optimization (PPO)-based discrete action introduced. Simulation results show that our DRL-based solutions outperform state-of-the-art algorithms. Moreover, PPO exhibits stable performance efficient compared benchmarks DQN Double (DDQN) strategies.