ERROR CONTROL AND ENERGY-CONSUMPTION IN COMMUNICATIONS FOR NOMADIC COMPUTING

We consider the problem of communications over a wireless channel in s upport of data transmissions from the perspective of small portable de vices that must rely on limited battery energy. We model the channel o utages as statistically correlated errors. Classic ARQ strategies are found to lead to a considerable waste of energy, due to the large numb er of transmissions. The use of finite energy sources in the face of d ependent channel errors leads to new protocol design criteria. As an e xample, a simple probing scheme, which slows down the transmission rat e when the channel is impaired, is shown to be more energy efficient, with a slight loss in throughput. A modified scheme that yields slight ly better performance but requires some additional complexity is also studied. Some references on the modeling of battery cells are discusse d to highlight the fact that battery charge capacity is strongly influ enced by the available ''relaxation time'' between current pulses. A f ormal approach that can track complex models for power sources, includ ing dynamic charge recovery, is also developed.