Multi agent PPO for terahertz cell free mobile edge computing networks
TL;DR
This work develops a multi-agent proximal policy optimization (MAPPO) algorithm to achieve a long-term optimal tradeoff between energy consumption and latency and demonstrates that under identical hyperparameter settings and training episodes, MAPPO achieves stable convergence and outperforms other MADRL baselines.
Abstract
Metaverse applications are emerging as a transformative technology for immersive experiences, yet their complex computational demands impose stringent latency and energy constraints. To address these challenges, we propose a Terahertz (THz) cell-free mobile edge computing (MEC) network. However, efficient joint task offloading and multi-dimensional resource allocation within this dynamic environment remain NP-hard problems. To tackle this, we develop a multi-agent proximal policy optimization (MAPPO) algorithm to achieve a long-term optimal tradeoff between energy consumption and latency. This multi-agent architecture effectively models the intricate coupling of decision variables, while tailored action space parameterization using Beta and Dirichlet distributions ensures strict adherence to physical constraints. Simulation results demonstrate that under identical hyperparameter settings and training episodes, MAPPO achieves stable convergence and outperforms other MADRL baselines, including MADDPG and MATD3. Furthermore, the proposed architecture significantly surpasses conventional cellular and small-cell systems.