Prediction based smart scheduler using neural network for wireless networks

Scheduling has been extensively studied in various disciplines such as operations research and wire-line networking. However, the unique characteristics of wireless networks, namely the time-varying channel conditions and multi-user diversity necessitate new scheduling solutions to be developed. The major focus of this paper is to propose a suitable scheduling algorithm, which can function effectively in a wireless heterogeneous environment. The scheduler gathers channel state information from the predictor and other service information from book-keeping available at each base station. Efficiency of scheduling algorithm greatly relies on predictor results. For this reason, channel prediction was performed using neural networks. Simulation reveals that, compared with the conventional one, the proposed scheduler is advantageous in terms of accuracy and speed. Performance comparison of channel predictors is also carried out in this paper.