Adaptive transmission with discrete code rates and channel state uncertainty

Without perfect channel state information at the transmitter, it is possible for adaptive transmission systems to experience information outage. In this paper, we formulate the throughput maximization with both an average power constraint and an information outage constraint. It is verified that, for the optimal transmission policy, the transmission only needs to adapt to a sufficient statistic for the channel state. For a Rayleigh fading channel with a simple training scheme, numerical results show that, with a reasonable amount of training and a small set of code rates, the adaptive transmission can achieve a performance very close to the ergodic capacity.