Reinforcement learning in power control games for internetwork interference mitigation in Wireless Body Area Networks

Publisher:
IEEE Xplore
Publication Type:
Conference Proceeding
Citation:
2012 International Symposium on Communications and Information Technologies, ISCIT 2012, 2012, pp. 256 - 262
Issue Date:
2012-12-01
Full metadata record
Files in This Item:
Filename Description Size
702973.pdfPublished version506.14 kB
Adobe PDF
A major cause of energy waste in wireless networks is the interference between nodes working in the same frequency band. This problem appears to be more serious in Wireless Body Area Networks (WBANs) in which energy is the most valuable resource. In order to cope with this issue, power control techniques can be employed. Amongst potential approaches, those with intelligence which allow WBANs to improve their performance by learning from experience can remarkably increase flexibility and adaptability. Besides, approaches with less inter-node negotiation and cooperation are more attractive in WBANBs due to their low overhead and superior scalability. In this paper, we propose a power controller which employs Reinforcement Learning (RL) to enable WBANs to learn from experience and coordinate their power levels in a distributed manner with no inter-node negotiation and cooperation. We evaluate the performance of the proposed power controller with different RL algorithms and compare them to a counterpart approach based on game theory without learning. Their performances are evaluated in terms of optimality of the solution and convergence. We show that the RL-based power controller can trade off throughput for transmission power level and achieve lower energy consumption compared to the counterpart game. © 2012 IEEE.
Please use this identifier to cite or link to this item: