A low-complexity congestion control and scheduling algorithm for multihop wireless networks with order-optimal per-flow delay

We consider the problem of designing a joint congestion control and scheduling algorithm for multihop wireless networks. The goal is to maximize the total utility and achieve low end-to-end delay simultaneously. Assume that there are M flows inside the network, and each flow m has a fixed route with H_m hops. Further, the network operates under the one-hop interference constraint. We develop a new congestion control and scheduling algorithm that combines a window-based flow control algorithm and a new distributed rate-based scheduling algorithm. For any ϵ, ϵ_m ∈ (0, 1), by appropriately choosing the number of backoff mini-slots for the scheduling algorithm and the window-size of flow m, our proposed algorithm can guarantee that each flow m achieves throughput no smaller than r_m(1 - ϵ)(1 - ϵ_m), where the total utility of the rate allocation vector r⃗ = [r_m] is no smaller than the total utility of any rate vector within half of the capacity region. Furthermore, the end-to-end delay of flow m can be upper bounded by H_m/(r_m(1 - ϵ)ϵ_m). Since a flow-m packet requires at least H_m time slots to reach the destination, the order of the per-flow delay upper bound is optimal with respect to the number of hops. To the best of our knowledge, this is the first fully-distributed joint congestion-control and scheduling algorithm that can guarantee order-optimal per-flow end-to-end delay and utilize close-to-half of the system capacity under the one-hop interference constraint. The throughput and delay bounds are proved by a novel stochastic dominance approach, which could be of independent value and be extended to general interference constraints. Our algorithm can be easily implemented in practice with a low per-node complexity that does not increase with the network size.