An Online Learning Method to Maximize Energy Efficiency of Cognitive Sensor Networks

In this letter, an online reinforcement learning method based on particle swarm optimization (PSO) is proposed to maximize transmission energy efficiency of cognitive radio sensor networks by regularizing packet lengths. The idea is to simulate a set of artificial channels to find a hypothetical channel that behaves almost equivalently to the actual channel by minimizing a properly developed loss function. This method eliminates the need for a separate offline channel estimation. The results show an improvement of, respectively, 40% and 20% in the energy efficiency of the proposed method compared with the constant packet length and MLE-based offline channel estimation. Also, the PSO-based optimization method outperforms similar evolutionary methods. This framework is general and can be used to optimize a desired parameter in cognitive radio networks.