Wireless Ad Hoc and Sensor Networks

Predictive Congestion Control for Wireless Sensor Networks 431In both CODA and Fusion protocols, nodes use a broadcast message toinform their neighboring nodes of the onset of congestion. Although thisis quite interesting, the onset of congestion message occurring inside thenetwork is not guaranteed to reach the sources. Moreover, available protocols(Hull et al. 2004, Wan et al. 2003, Yi and Shakkotai 2004) do notpredict the onset of congestion in dynamic environments, for example,due to fading channel conditions. Finally, very few analytical results arepresented in the literature in terms of guaranteeing the performance ofavailable congestion control protocols. In contrast to these protocols, themethod of Zawodniok and Jagannathan (2005) can predict and mitigatethe onset of congestion by gradually reducing the traffic flow defined byusing the queue availability and channel state.Besides predicting the onset of congestion, this scheme guarantees convergenceto the calculated target outgoing rate by using a novel, adaptivebackoff interval selection algorithm. In CSMA/CA-based wireless networks,a backoff selection mechanism is used to provide simultaneousaccess to a common transmission medium and to vary transmission rates.Many researchers (Vaidya et al. 2000, Wang et al. 2005, Kuo and Kuo 2003)have focused on the performance analysis of backoff selection schemesfor static environments. However, these schemes lack the ability to adaptto a changing channel state, congestion level, and size of the network. Forexample, in (Vaidya et al. 2000) the backoff intervals for the nodes areselected proportional to a flow weight, thus providing weighted fairness.However, the solution assumes a uniform density of transmitting nodesbecause the achieved throughput is determined by both a number ofcompeting nodes and their backoff intervals. Consequently, distributedfair scheduling (DFS) will yield an unfair throughput for nodes withdifferent number of neighbors. In contrast, the proposed algorithmdynamically alters backoff intervals according to current network conditions,for instance, the varying number of neighbor nodes and fadingchannels.Additionally, the protocol (Zawodniok and Jagannathan 2004) usesweights associated with flows to fairly allocate resources during congestion.By adding an optional dynamic weight adaptation algorithm,weighted fairness can be guaranteed in dynamic environments as shownin the proposed work. Finally, using a Lyapunov-based approach, thestability and convergence of the three algorithms, for buffer control, backoffinterval selection and dynamic weight adaptation, are proved.This chapter is organized as follows. Section 9.2 presents an overviewof the proposed methodology of predicting the onset of congestion, mitigation,DPC and weighted fairness. In Section 9.3, the flow control andbackoff interval selection schemes, their performance and weighted fairnessguarantees are presented. Mathematical analysis through theLyapunov method is discussed in this section. Section 9.4 details the