Ergodicity of Wireless Channels and Temporal Prediction

In this paper, we study the role of ergodicity in wireless channel prediction. Following the sinusoidal channel model, conditions under which the ergodic assumption is valid are presented. This sheds insight into when statistical channel models that employ ensemble averaging are appropriate. Due to the lack of ergodicity in a typical real world wireless channel, least squares prediction, an approach based on time averages is motivated as opposed to linear minimum mean squared error channel prediction, an approach based on ensemble averaging. We then study methods such as forward-backward and rank reduction for high quality channel prediction.