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