The end-to-end delay performance of a class of wireless scheduling algorithms

In this work, we are interested in designing scheduling algorithms for multihop wireless networks that can achieve good end-to-end delay performance. There has been evidence that some of the well-known queue-length-based throughputoptimal algorithms, such as the back-pressure algorithm, could have poor end-to-end delay performance. In particular, the end-to-end delay of the back-pressure algorithm may increase super-linearly with the number of hops due to the queue builtup. Assuming that the arrival rate is known and is inside the capacity region, in this talk we present a class of rate-based scheduling algorithms such that the end-to-end delay of each flow has an order dependent on the number of hops.