A framework for optimal decentralized power control with partial CSI

In this paper, we use a recent information theoretical result to develop a general framework for finding optimal power control policies in the case of interference channels. The aforementioned result characterizes the achievable payoffs for an N-Agent (transmitters in our application) coordination problem with a certain information structure. We then provide an algorithm which exploits the characterization of achievable payoffs by conditional probability distributions to find optimal decision functions for the transmitters. Due to its general nature, the developed framework is conducive for applications to diverse scenarios in wireless communications. In this article, we restrict our attention to the case of decentralized power control in interference channels for different utility functions namely sum-rate, sum-energy and sum-goodput. Our approach has the following salient features: 1) The method proposes optimal power control functions for any given utility function as opposed to ad-hoc solutions for different utilities proposed in the literature, and 2) Noise in the channel estimation is taken into account, thus providing robust optimal solutions.