IRS Assisted NOMA Aided Mobile Edge Computing With Queue Stability: Heterogeneous Multi-Agent Reinforcement Learning

By employing powerful edge servers for data processing, mobile computing (MEC) has been recognized as a promising technology to support emerging computation-intensive applications. Besides, non-orthogonal multiple access (NOMA)-aided MEC system can further enhance the spectral efficiency with massive tasks offloading. However, more dynamic devices brought online and uncontrollable stochastic channel environment, it is even desirable deploy appealing technique, i.e., intelligent reflecting surfaces (IRS), in flexibly tune communication environment improve energy efficiency. In this paper, we investigate joint offloading, computation resource allocation IRS-assisted NOMA system. We first formulate mixed integer maximization problem queue stability constraint. then propose Lyapunov-function-based Mixed Integer Deep Deterministic Policy Gradient (LMIDDPG) algorithm which based on centralized reinforcement learning (RL) framework. To be specific, design action space mapping contains both continuous mapping. Moreover, award function defined upper-bound of Lyapunov drift-plus-penalty function. enable end (EDs) choose actions independently at execution stage, Heterogeneous Multi-agent LMIDDPG (HMA-LMIDDPG) distributed RL framework homogeneous EDs heterogeneous base station (BS) multi-agent. Numerical results show that our proposed algorithms achieve superior performance benchmark while maintaining stability. Specially, structure HMA-LMIDDPG acquire gain than LMIDDPG.