A Hybrid Task Offloading and Resource Allocation Approach for Digital Twin-Empowered UAV-Assisted MEC Network Using Federated Reinforcement Learning for Future Wireless Network

Federated learning (FL) is proposed as a different approach for distributed learning on the edges while maintaining privacy. Existing FL methods, are mainly focused on learning deep classifying and clustering models, with little consideration offered to the federated reinforcement learning (FRL) tas...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on consumer electronics 2024-02, Vol.70 (1), p.3120-3130
Main Authors: Consul, Prakhar, Budhiraja, Ishan, Garg, Deepak, Kumar, Neeraj, Singh, Ramendra, Almogren, Ahmad S.
Format: Article
Language:English
Subjects:
Algorithms
Autonomous aerial vehicles
Biological system modeling
Clustering
Computation offloading
Computational modeling
Computer science
Consumer electronics
digital twin
Digital twins
Edge computing
Federated learning
federated reinforcement learning
Machine learning
Mobile computing
Mobile edge computing
Neural networks
Resource allocation
Servers
Task analysis
Task complexity
task offloading
unmanned aerial vehicle
Unmanned aerial vehicles
Wireless networks
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Federated learning (FL) is proposed as a different approach for distributed learning on the edges while maintaining privacy. Existing FL methods, are mainly focused on learning deep classifying and clustering models, with little consideration offered to the federated reinforcement learning (FRL) task on the edge, a difficult task in which several trained agents track local state and take local actions to train a global learning model without disclosing their local dataset. Several neural network models based on FRL have been presented recently to determine the best method for computation offloading (CO) and resource allocation (RA), particularly in Unmanned Aerial Vehicle (UAV) assisted Mobile Edge Computing (MEC). However, because of the complexity and variety of computational tasks involved in 6G and beyond networks, the FRL algorithms are challenging to apply directly to complicated UAV-assisted MEC scenarios. In this study, we present a generalized FRL approach based on a meta learning technique that incorporates RL models explained by numerous smart devices into a generic model. This research uses a normalized characteristic matrix to divide a complex network into small-scale units and provides a normalized network model for complex network situations based on the FRL meta critic method to determine the CO and RA strategy in a Digital Twin (DT)-enabled UAV-assisted MEC system. Numerical results show that the proposed scheme achieves higher and more reliable overall rewards. Proposed method achieves a 73.15% reduction in reward variance and a 14.23% increase in average rewards over 570 continuous operations.
ISSN:0098-3063
1558-4127
DOI:10.1109/TCE.2024.3368156