Wireless Ad Hoc and Sensor Networks

430 Wireless Ad Hoc and Sensor NetworksClosed-loop stability of the proposed hop-by-hop congestion control isdemonstrated by using the Lyapunov-based approach. Simulation resultsshow that this scheme results in fewer dropped packets, better fairnessindex, higher network efficiency and aggregate throughput, and smallerend-to-end delays over the other available schemes such as CongestionDetection and Avoidance (CODA) (Wan et al., 2003) and IEEE 802.11protocols.9.1 IntroductionNetwork congestion, which is quite common in wireless networks, occurswhen the offered load exceeds available capacity or the link bandwidthis reduced because of fading channels. Network congestion causes channelquality to degrade and loss rates to rise. It leads to packet drops atthe buffers, increased delays, and wasted energy, and requires retransmissions.Moreover, traffic flow will be unfair for nodes whose data have totraverse a significant number of hops. This considerably reduces the performanceand lifetime of the network. Additionally, WSN have constraintsimposed on energy, memory, and bandwidth. Therefore, energy-efficientdata transmission protocols are required to mitigate congestion resultingfrom fading channels and excess load. In particular, a congestion controlmechanism is needed to balance the load, to prevent packet drops, andto avoid network deadlock.Rigorous work has been done in wired networks on end-to-end congestioncontrol (Jagannathan 2002, Peng et al. 2006). In spite of severaladvantages in end-to-end control schemes, the need to propagate the onsetof congestion between end-systems makes the approach slow. In general,a hop-by-hop congestion control scheme reacts to congestion faster andis normally preferred to minimize packet losses in wireless networks.Therefore, the scheme from Zawodniok and Jagannathan (2005) uses anovel hop-by-hop flow control algorithm that is capable of predicting theonset of congestion and then gradually reducing the incoming traffic bymeans of a backpressure signal.In comparison, the CODA protocol proposed by Wan et al. (2003) usesboth a hop-by-hop and an end-to-end congestion control scheme to reactto an already present congestion by simply dropping packets at the nodepreceding the congestion area. Thus, CODA partially minimizes theeffects of congestion, and as a result retransmissions still occur. Similar toCODA, Fusion (Hull et al., 2004) uses a static threshold value for detectingthe onset of congestion even though it is normally difficult to determinea suitable threshold value that works in dynamic channel environments.