Research Article
Q-Learning Based Optimum Relay Selection fora SWIPT-Enabled Wireless System
@INPROCEEDINGS{10.1007/978-3-030-90196-7_10, author={Haojie Wang and Bo Li}, title={Q-Learning Based Optimum Relay Selection fora SWIPT-Enabled Wireless System}, proceedings={Artificial Intelligence for Communications and Networks. Third EAI International Conference, AICON 2021, Xining, China, October 23--24, 2021, Proceedings, Part I}, proceedings_a={AICON}, year={2021}, month={11}, keywords={Simultaneous wireless information and power transfer Reinforcement learning Relay selection}, doi={10.1007/978-3-030-90196-7_10} }- Haojie Wang
Bo Li
Year: 2021
Q-Learning Based Optimum Relay Selection fora SWIPT-Enabled Wireless System
AICON
Springer
DOI: 10.1007/978-3-030-90196-7_10
Abstract
This paper exploits q-learning to solve the optimization of relay resource allocation in a SWIPT system in order to maximize the overall source nodes communication rate. The traditional method based on greedy strategy search can find the current optimal relay. However, this method does not adapt to the changes of the dynamic network and is easy to fall into the local optimum. In q-learning, agents are driven to intelligently switch strategies to obtain rewards for complex dynamic environments. A multi-source node competition relay environment is created. Our proposed scheme treats the optimised objective as an environmental reward and models the process of relay selection as a Markovian decision process. After experimental simulations and comparisons, the proposed scheme improves the overall throughput and resource utilisation in an environment where multiple source nodes compete with each other, and performs significantly better than the greedy strategy.
