Capacity of cognitive channel and power allocation

This paper addresses power allocation between channels in a cognitive radio system, and uses insights from information theory to compare the total capacity available using a fully distributed power allocation scheme with an optimum scheme based on perfect knowledge of all links. First the equations for optimum allocation are obtained, leading to an extension of the water-filling power allocation, which can be solved iteratively. Then algorithms which can be implemented on a distributed basis are considered, and compared with the capacity of the optimum scheme, using a channel model including both shadow and Rayleigh fading.