SIR Prediction for Downlink Packet Access

In cellular systems, with frequency reuse of one, basestations strongly interfere with each other on downlink. Typically more than forty percent of the cell area faces a signal to interference ratio of 0 dB or less. To improve system throughput, packet transmissions are scheduled by the base when signal to interference ratio is higher than each mobile´s average. Downlink pilots are measured and the SIR is fed back as a measure of the achievable rate. We examine prediction of the channel signal to interference ratio by the mobile. Both powers are modeled as auto-regressive stochastic processes. We derive the probability distribution function of the prediction error to determine the backoff margin required for a percentage of packets to be decoded successfully. The performance improvement due to prediction for a proportional fair scheduled system is shown through simulation