Paper Info
Title
Online Policies For Throughput Maximization Of Backscatter Assisted Wireless Powered Communication Via Reinforcement Learning Approaches
Abstract
In this paper, we consider the design of online policies in a backscatter assisted wireless powered communication system, aiming at maximizing the longterm average throughput. We consider a complete data life cycle, from sampling, compression, transmission to reception and decompression. Practical constraints including finite battery capacity, time-varying uplink channel and nonlinear energy harvesting model are considered. An optimization problem is formulated in a Markov decision process framework to maximize the longterm average throughput by a hybrid of mode switching, time and power allocation, and compression ratio selection. Capitalizing on this, we first adopt value iteration (VI) algorithm to find offline optimal solution as benchmark. Then, we propose Q-learning (QL) and deep Q-learning (DQL) algorithms to obtain online solutions without prior information. Simulation results demonstrate the effectiveness of the hybrid transmission mode with flexible data compression. Furthermore, DQL-based online solution performs the closest to the optimal VI-based offline solution and significantly outperforms the other two baseline schemes QL and random policy. Insight analysis on the structure of the optimal policy is also provided. (C) 2021 Elsevier B.V. All rights reserved.
Year
DOI
Venue
2021
10.1016/j.pmcj.2021.101463
PERVASIVE AND MOBILE COMPUTING
Keywords
DocType
Volume
Wireless powered communication, Backscatter communication, Longterm average throughput, Markov decision process, Deep Q-learning
Journal
77
ISSN
Citations 
PageRank 
1574-1192
0
0.34
References 
Authors
0
6
Name
Order
Citations
PageRank
Xiaofeng Su100.34
Yanjun Li28725.03
Meihui Gao300.34
Zhibo Wang478679.49
Yinglong Li500.34
Yi-hua Zhu6287.45