Optimal Control of Data Transmission over a Fluctuating Channel with Unknown State

Optimization problem for data packet transmission over a communication channel governed by a hidden Markov process is considered. The transmitter is modeled as a single-channel finite-buffer queuing system with non-stationary Poisson arrivals. The service rate is proportional to the controlled transmission rate with channel-dependent factor. Buffer overflow leads to packet losses, whereas channel state worsening results in lower service rate. The goal of the optimization problem is to minimize average losses under constraint on the transmitter energy consumption. The exact form of the optimal policy is presented for the augmented control problem. Several control policies with incomplete information are proposed on the basis of the optimal control and hidden state estimates. We consider two estimates based on the optimal filtering equations and the current queue state. Results of computer simulation are presented to compare the control policies under consideration.