Author : Lu Jin
Publisher :
ISBN 13 :
Total Pages : 91 pages
Book Rating : 4.:/5 (93 download)
Book Synopsis Reinforcement Learning Based Energy Efficient Routing Protocols for Underwater Acoustic Wireless Sensor Networks by : Lu Jin
Download or read book Reinforcement Learning Based Energy Efficient Routing Protocols for Underwater Acoustic Wireless Sensor Networks written by Lu Jin and published by . This book was released on 2012 with total page 91 pages. Available in PDF, EPUB and Kindle. Book excerpt: [Truncated abstract] The unique properties of underwater acoustic communications, such as large and time-varying propagation delay, low and range dependent channel bandwidth, and adverse operating environments, challenge the design of an energy efficient protocol in underwater acoustic wireless sensor networks (UA-WSNs). In this dissertation, we present a family of reinforcement learning based energy-efficient protocols for UA-WSNs so that the systems are "self-adaptive" or "self-optimized" to the environment after having been deployed. We show that, due to the limited bandwidth in underwater acoustic communication channels, the energy consumption caused by collisions and retransmissions is much higher than that of its terrestrial counterparts. Therefore, the key issue of extending the lifetime of underwater sensor networks is to reduce the collisions and retransmissions. To address this issue in a single underwater acoustic channel, we model the UA-WSNs as single agent reinforcement learning systems and design an routing protocol to aid transmitters to determine the most energy efficient relay nodes. In the design of the reward function, we use the transmission power level, the neighborhood interference, and the feasible bandwidth as the main parameters for the selection of relay nodes. Extensive simulation results demonstrate that the proposed protocol achieves a remarkable energy efficiency in comparison with the Ad hoc On-Demand Distance Vector (AODV) routing protocol. For verifying the feasibility of the reinforcement learning based energy efficient protocol in multiple underwater acoustic communication channels, we improve the protocol by including the channel-related factors in the design of the reward function. In the proposed UA-WSN system, the feasible acoustic spectrum is divided into a number of sub-channels which suffer diverse and frequency dependant attenuation and noise interference. The proposed protocol helps each transmitter to determine not only an optimal relay node, but also an optimal propagation sub-channel to the optimal relay node. For the reward function of the proposed protocol, the channel-related factors are used to evaluate the quality of the sub-channels in terms of noise and attenuation level, probability of collisions, etc.. Simulation results show that the use of a multiple-channel reinforcement learning protocol manifests far better energy efficiency than a single-channel reinforcement learning protocol. But the use of a large number of sub-channels leads to performance degradation due to the extremely narrow bandwidth and low capacity of sub-channels.