Suhas N. Diggavi, Fellow, IEEESeptember 22nd, 2016
FiledSeptember 22nd, 2016It is worth mentioning that the power optimization problem stated in Theorem 2 is a non-convex problem. Although the closed-form solution of the this problem is not easy to derive explicitly, but by using dynamic programming it can be easily found numerically. In Section VII, based on the structure of this optimization problem and by exploiting special properties of its KKT necessary conditions for the optimality, we propose a dynamic programming algorithm that finds the optimal power allocation (see Algorithms 1 and 2). More specifically, we have the following theorem.