Interference estimation with noisy measurements in broadband wireless packet networks

It has been shown that link adaptation and power control can improve the performance of our future wireless packet networks. Realizing the expected performance gain of these techniques requires accurate prediction of future interference power. We propose a new method based on Kalman filtering for interference estimation. The new method is devised by observing: (a) it is possible to identify fairly accurately the number of active co-channel interferers in the cellular networks and (b) interference power is positively correlated with the number of active interferers. The new technique uses a two-dimensional Kalman filter to exploit that correlation to enhance the prediction accuracy. Using a cellular network with 1/3 frequency reuse and partial traffic loading, the performance of the new method is compared with a simplified method using a one-dimensional Kalman filter where the number of active interferers is not considered. Further, the new method is compared with the traditional exponential filtering. Since the proposed method and the simplified method track interference and measurement errors separately, their predictions represent closely the best estimation by exponential filtering with the optimal parameter. In addition, for a typical network environment, the two-dimensional method yields the lowest prediction errors for a wide range of parameters, and provides a 0.5 dB improvement for the 90th percentile estimation error over the simplified method due to exploitation of the positive correlation between interference and the number of active interferers