# DRL-based Distributed Resource Allocation for Edge Computing in Cell-Free Massive MIMO Network

**Authors:** Fitsum Debebe Tilahun, Ameha Tsegaye Abebe, Chung G. Kang  
**arXiv:** [2208.10725](https://arxiv.org/abs/2208.10725)  
**Date:** 2022-08-23  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2208.10725v1/

## Key points

- Authors address the need for advanced multimedia services and meeting QoS requirements
- Proposal to leverage cell-free massive MIMO-enabled mobile edge network
- Advantage of reliable cell-free links in offloading computation to edge server
- Formulation of joint communication and computing resource allocation problem
- Proposed solution approach based on cooperative multi-agent reinforcement learning framework
- Simulation results show superiority over heuristic baselines and convergence to centralized benchmark
- Comparison with cellular Mobile Edge Computing (MEC) systems, showing better performance in cell-free system
- Innovative approach for addressing computing and QoS requirements of advanced multimedia services
- DRL-based distributed resource allocation algorithm demonstrates improved performance while minimizing energy consumption

## Summary

In this paper titled "DRL-based Distributed Resource Allocation for Edge Computing in Cell-Free Massive MIMO Network," authors Fitsum Debebe Tilahun, Ameha Tsegaye Abebe, and Chung G. Kang address the growing demand for advanced multimedia services and the need to meet stringent computing and quality-of-service (QoS) requirements. They propose a solution that leverages the benefits of a cell-free massive MIMO-enabled mobile edge network. The authors highlight the advantage of reliable cell-free links in offloading intensive computation to the edge server, allowing resource-constrained end-users to enhance their on-board processing capabilities with edge computing. To optimize energy consumption while meeting user-specific deadlines, they formulate a joint communication and computing resource allocation (JCCRA) problem. To tackle this problem, the authors propose a fully distributed solution approach based on a cooperative multi-agent reinforcement learning framework. Each user is implemented as a learning agent that makes joint resource allocation decisions based solely on local information. The simulation results demonstrate that their proposed distributed approach outperforms heuristic baselines and converges to a centralized target benchmark without incurring significant overhead. Furthermore, the authors compare their algorithm's performance in a cell-free system with cellular Mobile Edge Computing (MEC) systems such as small cell-based MEC systems. They show that their proposed algorithm performs significantly better in the cell free system compared to other approaches. Overall, this paper presents an innovative approach for addressing computing and QoS requirements of advanced multimedia services by leveraging cell free massive MIMO enabled mobile edge networks. The proposed DRL based distributed resource allocation algorithm demonstrates improved performance compared to baseline methods while minimizing energy consumption for users in these networks.