Temporal difference method-based multi-step ahead prediction of long term deep fading in mobile networks

Abstract

In this paper, the problem of multi-step ahead prediction of long term deep fading in mobile networks is studied. We first briefly discuss the operating principle of the temporal difference (TD) method. A TD method-based multi-step ahead prediction scheme using the modified Elman neural network (MENN) is then proposed. This prediction approach provides for on-line adaptation and fast convergence rate. Next, it is applied to the prediction of the occurrence of long term deep fading in the mobile communications systems. Simulation experiments reveal that our prediction scheme is capable of predicting the degree of occurrence possibility of future deep fading. The prediction results are considered to be a solid basis for employing the reinforcement learning method in the power control of cellular phone systems.