Stochastic power control with bit estimation and base station processing

A distributed power control algorithm for a stochastic system when the estimates of the measured quantities are noisy has been proposed in Ululkus and Yates(1998). In this paper, we propose two different modifications of this algorithm that aim at lowering the noise variance of the measured quantities in order to achieve a faster rate of convergence. In the first approach, the received (possibly erroneous) bit of the desired user is used to obtain a more accurate estimate of its interference power. While in the second approach, the local information available at each base station is used to simulate an estimate of the interference power experienced by each one of its own users. Then a weighted average of this simulated interference power and the actual measured one is taken, effectively resulting in an average over a larger window, which lowers the noise variance. We demonstrate through simulations that both of these modifications achieve a faster rate of convergence compared to that of the existing algorithms for a distributed and stochastic system.