Practical Implementation Issues of Reinforcement Learning Based ALOHA for Wireless Sensor Networks
Conference: ISWCS 2013 - The Tenth International Symposium on Wireless Communication Systems
08/27/2013 - 08/30/2013 at Ilmenau, Deutschland
Proceedings: ISWCS 2013
Pages: 5Language: englishTyp: PDF
Personal VDE Members are entitled to a 10% discount on this title
Authors:
Kosunalp, Selahattin; Mitchell, Paul D.; Grace, David; Clarke, Tim (Communications Research Group, Department of Electronics, University of York York, United Kingdom)
Abstract:
This paper presents a practical implementation of the Reinforcement Learning (RL) based ALOHA protocols (RLALOHA) for single-hop communication. The practical implementation issues of RL-ALOHA are studied. A potential real-world phenomenon that impacts on performance is brought to light and a new scheme is proposed to deal with the phenomenon which provides the ideal throughput performance of RL-ALOHA. The practical results show that the proposed scheme improves the convergence properties to the steady state of one slot assigned per node.