Bandit-based relay selection in cooperative networks over unknown stationary channels
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Standard
Bandit-based relay selection in cooperative networks over unknown stationary channels. / Nomikos, Nikolaos; Talebi, Sadegh; Wichman, Risto; Charalambous, Themistoklis.
Proceedings of the 2020 IEEE 30th International Workshop on Machine Learning for Signal Processing, MLSP 2020. IEEE, 2020. 9231604.Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Bandit-based relay selection in cooperative networks over unknown stationary channels
AU - Nomikos, Nikolaos
AU - Talebi, Sadegh
AU - Wichman, Risto
AU - Charalambous, Themistoklis
PY - 2020
Y1 - 2020
N2 - In recent years, wireless node density has increased rapidly, as more base stations, users, and machines coexist. Exploiting this node density, cooperative relaying has been deployed to improve connectivity throughout the network. Such a configuration, however, often demands relay scheduling, which comes with increased channel estimation and signaling overheads. To reduce these overheads, in this paper, we propose low-complexity relay scheduling mechanisms with the aid of a multi-armed bandit (MAB) framework. More specifically, this MAB framework is used for relay scheduling, based only on observing the acknowledgements/negative-acknow-ledgements (ACK/NACK) of packet transmissions. Hence, a bandit-based opportunistic relay selection (BB - ORS) mechanism is developed, recovering eventually the performance of classical opportunistic relay selection (0RS) when channel state information (CSI) is available without requiring any CSI. In addition, a distributed implementation of BB - ORS is presented, herein called d - BB - ORS, where distributed timers are used at the relays for relay selection, thus reducing the signaling overhead significantly. BB - ORS is compared to optimal scheduling with full CSI and the negligible performance gap is compensated by the low-complexity low-overhead implementation, while it surpasses the performance of ORS with outdated CSI.
AB - In recent years, wireless node density has increased rapidly, as more base stations, users, and machines coexist. Exploiting this node density, cooperative relaying has been deployed to improve connectivity throughout the network. Such a configuration, however, often demands relay scheduling, which comes with increased channel estimation and signaling overheads. To reduce these overheads, in this paper, we propose low-complexity relay scheduling mechanisms with the aid of a multi-armed bandit (MAB) framework. More specifically, this MAB framework is used for relay scheduling, based only on observing the acknowledgements/negative-acknow-ledgements (ACK/NACK) of packet transmissions. Hence, a bandit-based opportunistic relay selection (BB - ORS) mechanism is developed, recovering eventually the performance of classical opportunistic relay selection (0RS) when channel state information (CSI) is available without requiring any CSI. In addition, a distributed implementation of BB - ORS is presented, herein called d - BB - ORS, where distributed timers are used at the relays for relay selection, thus reducing the signaling overhead significantly. BB - ORS is compared to optimal scheduling with full CSI and the negligible performance gap is compensated by the low-complexity low-overhead implementation, while it surpasses the performance of ORS with outdated CSI.
KW - Machine learning
KW - Multi-armed bandits
KW - Relay selection
KW - Upper confidence bound policies
U2 - 10.1109/MLSP49062.2020.9231604
DO - 10.1109/MLSP49062.2020.9231604
M3 - Article in proceedings
AN - SCOPUS:85096503271
BT - Proceedings of the 2020 IEEE 30th International Workshop on Machine Learning for Signal Processing, MLSP 2020
PB - IEEE
T2 - 30th IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2020
Y2 - 21 September 2020 through 24 September 2020
ER -
ID: 255502821