|
Decentralized Reinforcement Learning for Wake-up Scheduling Host Publication: European Conference on Wireless Sensor Networks Authors: M. Emilov Mihaylov, Y. Le Borgne, A. Nowé and K. Tuyls Publisher: University of Coimbra Publication Date: Feb. 2010 Number of Pages: 3 ISBN: 978-989-96001-3-3
Abstract: Latency is a key issue in the application of delay-critical
sensor nets. Reducing latency requires non trivial synchronization of the
sensor nodes. This work presents a decentralized reinforcement learning
algorithm allowing to decrease the end-to-end latency. Each node learns
on-line the most benecial period to stay awake, such that it avoids
idle listening and overhearing. We illustrate this approach and using the
OMNET++ simulator we show that it reduces end-to-end latency in a
network while preserving its energy efficiency.
|
