Search

In multihop ad hoc networks over 802.11, the real bandwidth-delay product (BDP) cannot reach a value as large as in wired networks due to contention through wireless shared channels. As a result, conventional congestion-control algorithms cause an overshooting window problem and result in poor throughput in ad hoc networks. In this paper, we split the real round-trip time (RTT) into two parts: 1) congestion RTT and 2) contention RTT. We reveal that the contention RTT has nothing to do with the BDP and that the BDP is determined by only the congestion RTT if a link with the worst contention status does not lead to link breakage. An inadequate use of contention RTT causes a Transmission Control Protocol (TCP) congestion window overshooting problem. We propose a novel mechanism called congestion window adaptation through contention detection, which provides a more accurate method of estimating the contention status. Based on this mechanism, we present a congestion window adaptation method to limit the window size from overshooting. Simulation results show that our proposed mechanism outperforms the conventional TCP and the enhanced mechanisms called TCP with contention control and TCP with a maximum window. [ABSTRACT FROM PUBLISHER]

In this paper, we propose two cross-layer algorithms, namely, the Power-optimal Scheduling Algorithm (PSA) and the Throughput-optimal Scheduling Algorithm (TSA), to minimize energy consumption and to maximize throughput, respectively, in multihop wireless networks. Our algorithms guarantee a flow-based minimum data rate and jointly integrate congestion control, power allocation, routing, and link rate scheduling. Different from traditional algorithms that assume infinite buffers, the proposed algorithms deterministically upper bound the flow-based packet queue length and thus can be employed in multihop networks with finite buffers. In addition, the algorithms achieve a power expenditure/throughput “\epsilon-close” to the optimal value, with a tradeoff of order O(\1/\epsilon) in buffer size. The average end-to-end delay upper bound can also be derived from the finite buffer property. Finally, numerical results are presented to show the performance of the two algorithms with different system parameters. [ABSTRACT FROM PUBLISHER]