Title: Combined Channel Estimation and Power Control for Wireless Fading Channels
Speaker: Professor Michael Honig
Speaker Info: Northwestern University
One of the main challenges to achieving reliable digital communications over a mobile wireless link is channel fading, which causes random variations in the amplitude and phase of the received signal. Knowledge of the fading process is useful for estimating the transmitted symbols at the receiver, and for controlling the transmit power to compensate for variations in the received signal strength.Date: Thursday, January 25, 2007
We consider a communications model in which neither the receiver nor the transmitter knows the instantaneous fading process, which is assumed to be complex Gaussian. Based on the fading statistics, the receiver estimates the fading process through the use of known transmitted (training) symbols, and the current estimate is fed back to the transmitter for power control. The problem is then to allocate the transmitted power across training and data (information-bearing) symbols to maximize the achievable information rate. We study a continuous limit, in which the channel coherence time tends to zero, so that the fading process becomes a diffusion. The optimal power allocation is then determined by the corresponding limit of a Bellman equation. In this limit we show that the optimal training policy is "bang-bang", i.e., the training power switches between zero and a maximum value, depending on the system state. Open issues and directions for further work will be discussed.
(Joint work with M. Agarwal and B. Ata.)