Adaptive transmission of variable-rate data over a fading channel for energy-efficiency

The paper explores the adaptation of transmission rate and power jointly to the data generation rate and channel fading, for minimizing transmission energy. The optimal offline adaptation problem is solved, which provides a lower-bound on the transmission energy consumed by any practical, that is, online, scheme. A heuristic online algorithm, look-ahead water-filling, is developed for adapting to the queue state as well as the channel state, and is shown through simulations to achieve transmission energy per packet close to optimal. As the packet arrival rate is varied within known limits, the average energy per packet used by look-ahead water-filling is significantly lower than that achieved by optimal adaptation to the channel only (water-filling in time). The delay per packet is larger, but is almost constant for all data arrival rates. The results can be generalized to multi-access and broadcast fading channels.