Implementing backpressure-based rate control in wireless networks

From a theoretical standpoint, backpressure-based techniques present elegant cross-layer rate control solutions that use only local queue information. It is only recently that attempts are being made to design real world wireless protocols using these techniques. To aid this effort, we undertake a comprehensive experimental evaluation of backpressure mechanisms for multihop wireless networks, in particular, the first such study in the context of wireless sensor networks. Our evaluation yields two key insights into the design of such protocols. First, for wireless sensor networks, we show that a simple backpressure scheduling policy which allows nodes to transmit so long as they have a positive queue differential (irrespective of its size) gives performance comparable to more sophisticated heuristics. The advantage of this approach is that no changes are required to the underlying MAC. Second, we show that the performance of backpressure based protocols is highly sensitive to a parameter setting that depends upon current traffic conditions. Therefore, practical backpressure protocols must provide for automatic parameter adaptation.