A fuzzy-based function approximation technique for reinforcement learning1

Wu, Cheng; Song, Huichun; Yan, Changsheng; Wang, Yiming

doi:10.3233/ifs-162212

Cited by 7 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared to Q-learning, Sarsa-learning can select the action to follow the strategy and update the action value function to follow the same strategy. Q-learning way can simplify the execution of algorithm analysis and convergence difficulty, but Sarsa-learning has a higher learning efficiency and faster convergence rate [50]. The algorithm employs a simple update process for value iteration.…”

Section: Sarsa: Online Control For Cwsnmentioning

confidence: 99%

Energy-efficiency opportunistic spectrum allocation in cognitive wireless sensor network

Wang

Yin

2018

J Wireless Com Network

Self Cite

View full text Add to dashboard Cite

The developments in wireless sensor network (WSN) that enriches with the unique capabilities of cognitive radio technique are giving impetus to the evolution of Cognitive Wireless Sensor Network (CWSN). In a CWSN, wireless sensor nodes can opportunistically transmit on vacant licensed frequencies and operate under a strict interference avoidance policy with the other licensed users. However, typical constraints of energy conservation from batterydriven design, local spectrum availability, reachability with other sensor nodes, and large-scale network architecture with complex topology are factors that maintain an acceptable network performance in the design of CWSN. In addition, the distributed nature of sensor networks also forces each sensor node to act cooperatively for a goal of maximizing the performance of overall network. The desirable features of CWSN make Multi-agent Reinforcement Learning (RL) technique an attractive choice. In this paper, we propose a reinforcement learning-based transmission power and spectrum selection scheme that allows individual sensors to adapt and learn from their past choices and those of their neighbors. Our proposed scheme is multi-agent distributed and is adaptive to both the end-to-end source to sink data requirements and the level of residual energy contained within the sensors in the network. Results show significant improvement in network lifetime when compared with greedy-based resource allocation schemes.

show abstract

Section: Sarsa: Online Control For Cwsnmentioning

confidence: 99%