Skip to Main Content
In an energy-efficient wireless communication system, transmit powers are minimized subject to predetermined signal-to-interference ratio (SIR) requirements. In this paper, a general framework for distributed stochastic power control (PC) algorithms is proposed, where the transmit powers are updated based on stochastic approximations. The proposed algorithms are distributed in the sense that no global information is needed in the power updates. Interference to each user is estimated locally via noisy observations. Two types of stochastic PC algorithms are studied: standard stochastic PC algorithms where the interference estimator is unbiased, and quasi-standard stochastic PC algorithms where the interference estimator is biased. The conditions under which the stochastic PC algorithms converge to the unique optimal solution are identified. Corresponding to two classes of iteration step-size sequences, two types of convergence, the probability one convergence and convergence in probability, are shown for both algorithms based on recent results in the stochastic approximation literature. Based on the theoretical results, some well-known stochastic PC algorithms, such as stochastic PC with matched filter receivers, and joint stochastic PC with blind minimum mean-squared error (MMSE) interference suppression, are revisited; several new stochastic PC algorithms, such as stochastic PC with minimum-power base-station assignment, and stochastic PC with limited diversity, are proposed. It is shown that these algorithms fall into either the standard or the quasi-standard stochastic PC framework. Simulation results are given to illustrate the performance of the proposed algorithms in practical systems.