SIR PREDICTION FOR DOWNLINK PACKET ACCESS
23 October 2006
Downlink cellular packet systems with opportunistic scheduling and adaptive modulation must deal with outdated channel state information. Typically, a fade margin is included to mitigate the effects of channel variation due to the delay and processing in the feed back channel. Channel prediction can provide improvement, but the fade margin must remain to account for prediction errors. The optimum fade margin is derived to obtain a certain outage for any prediction error variance. In cellular systems the frequency reuse is often one, and basestations will interfere strongly on downlink. It may be necessary then to predict the signal as well as a small number of dominate interferers. These powers are modeled as autoregressive stochastic processes, and the optimum MAP predictor is derived. Utilizing our method for determining the fade margin, conservative rate estimates of each user are made, and improvement in throughput is shown compared to a proportional fair scheduled system without prediction.
