Deep learning approach on channel selection strategy for minimizing co-channel interference in unlicensed channels
Date
2020
Authors
Yen, L.
Adege, A.B.
Lin, H.P.
Ho, C.H.
Lever, K.
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Journal article
Citation
Microelectronics Reliability, 2020; 105(article no. 113558):1-12
Statement of Responsibility
Conference Name
Abstract
Without centralized frequency reuse coordination, unlicensed Wi-Fi spectrum users suffer from strong co-channel interference. The conventional Wi-Fi spectrum co-existence mechanism, carrier-sense multiple access with collision avoidance, temporarilly pauses the packet transmission for co-channel interference but still raises ambient noise level. In this paper, we propose the channel selection strategies for unlicensed Wi-Fi users in both the static case and the moving case. For static users, the strategy is to use channel switching implemented on a commercial off-the-shelf access point. This strategy senses the situation of current channel usage and searches for an idle channel if the currently used channel is occupied by other acess points. The designed strategy then manipulates the access point to switch to the idle channel. Experimental results show that the designed channel switching strategy outperforms the carrier-sense multiple access with collision avoidance with about 88% increment of the data throughput measured. However, the moving users may not accept the service outage due to the hardware channel switching. Therefore, a novel design of prediction based channel selection strategy without channel switching using deep learning is proposed, including the deep learning based channel prediction strategy and the time-to-live based algorithm. At each point on the trajectory of the moving user, the presented prediction strategy firstly predicts the three channels that have the strongest interference power with the deep neural network. The presented algorithm then rules out these interference-rich channels. Finally, a channel with lower interference can be found and selected for the moving users. Performance evaluation results show that along the trajectory, in compared with the conventional selection of non-overlapping channel, the presented channel selection strategy decreases up to about 36 dB in the interference power level and enhances up to about 37 dB in signal-to-interference-plus-noise ratio.
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
Copyright 2019 Elsevier