FedMAPPO: Federated Multi-Agent Proximal Policy Optimization for Energy-Efficient Task Offloading in Industrial IoT Networks
Abstract
Abstract In an Industrial Internet of Things (IIoT) network, billions of devices with sensors harvest huge streams of data that must be processed with strict latency requirements and low energy consumption, a fundamental issue is the task offloading which should be handled in an en-ergy efficient and privacy-aware manner [3, 5]. In this paper, to enable energy-efficient task offloading in IIoT, we introduce FedMAPPO, a Federated Multi-Agent Deep Reinforcement Learning (DRL) framework, which synergistically combines Multi-Agent Proximal Policy Optimization (MAPPO) [20, 23] and Federated Learning (FedAvg) [24] under the CTDE paradigm. We conduct rigorous experiments, where we run 30 independent simulation runs for each method (210 simulation runs in total, including 7 methods: 4 conventional base-line methods (Cloud-Only, Edge-Only, Threshold-Based and Lyapunov [16]) and 3 DRL-based methods (DQN Baseline [17, 18], MAPPO without FL [21, 22] and the proposed FedMAPPO). Experimental results indicate that FedMAPPO can achieve energy reduction of 99.99% compared to Cloud-Only processing (407,685 J vs. 7.05 billion J), 99.98% energy re-duction compared to the DQN Baseline, and 99.99% energy reduction compared to MAPPO without FL while keeping the task success rate of 68.89% and the minimum battery level of 95.42%. The improvement is statistically significant ( p 0 . 001, paired t -test) and small 95% confidence intervals for all improvements. The unique contribution of the framework is a gain…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.