Energy efficient wireless scheduling: adaptive loading in time

When designing wireless systems, one of the major challenges is tackling the time depending fading behavior of the channel. To achieve maximum throughput under a power constraint, techniques have been proposed which adapt the modulation on the fly, based on the instantaneous channel condition. However, when the design goal is minimizing the overall energy consumption under a throughput constraint, we have to tackle a more complex scheduling problem. The reason is that energy optimality might require deliberately decreasing the transmission rate at times, if we know that the rate loss can be compensated for in the future when channel conditions are more favorable. We present a solution to the problem of minimum energy scheduling on wireless links by exploiting an analogy with the adaptive bit loading problem in multicarrier systems. Instead of allocating bits across multiple channels with different quality, we formulate the problem as one of allocating bits at the different time instants. The analogy is however not straightforward because one does not know the future, and therefore cannot simply apply conventional bit loading to the time dimension. We devise a new technique that approximates adaptive loading in time, but only depends on the instantaneous channel condition. Our algorithm is simple to implement, and shows up to 5× reduction in energy over existing approaches.