Buffer and channel adaptive modulation for transmission over fading channels

We consider the problem of adaptive modulation for increasing the throughput of transmission over fading channels. In the model, packets arrive at a finite-length buffer according to a Poisson distribution and are mapped into M-ary quadrature amplitude modulated (MQAM) symbols for transmission over a correlated Rayleigh fading channel. We assume that buffer and channel state information are always available at the transmitter. Our objective is to vary the transmit power and the signal constellation size of the modulator according to both the buffer and channel states so that the system throughput is maximized under some average transmit power and bit error rate (BER) constraints. We formulate this optimization problem as a Markov decision process (MDP) and use dynamic programming techniques to obtain the solution. More importantly, we show that, under certain conditions, the optimal transmission rate increases when the channel gain decreases toward the outage threshold - the point below which communication is not possible. This is in contrast to the water-filling structure of the link adaptive policy that achieves capacity on fading channels.