Research On The Congestion Control Of Wireless Sensor Network

Wireless Sensor Network(WSN) have node density, poor capacity of data processing andstoring, and the feature of "many-to-one" data gathering, coupled with fierce competition ofchannel resources and the unstable of link quality of wireless network, which all cause thecongestion problems of WSN become very complex and serious.Congestion causes many problems. It increases the delay of packet transmission,decreases the throughput, causes the packet loss, and even results in network paralysis. TCP, atransport layer protocol, is responsible for the network congestion in traditional networks. Inwhich congestion is main reason for packet loss, so TCP judges whether the network iscongested by packet loss or not. However, in WSN, in addition to network congestion, routingfracture, poor quality of the channel and mutual interference of sudden data flow are all likelyto be the cause of packet loss. The packet loss due to "non-congested" events will promote theimplementation of the congestion control strategy when TCP is applied directly to wirelesssensor networks, which decreases overall network performance. In addition, the rapidvariation of network conditions and channel competition result in information failure beforethe destination node receive it, with end-to-end congestion notification program. Therefore, ithas become the key technology for wireless sensor networks to study on efficient congestioncontrol.The main work of this paper is improve the performance of network congestion controlmechanisms for wireless sensor network，respectively from the following three aspects:1) Transport layer: Similar with TCP, DCCP doesn’t distinguish the reason of packetloss between wireless link error and network congestion, which results in poor networkthroughput performance; Secondly, congestion detection mechanism based on packet loss cannot accurately reflect the state of the network. So, the new mechanism differentiates betweenwireless link errors and network congestion in terms of packet loss rate, at the same time, thechannel busyness ratio is introduced to collaborative detection of network congestion, anddifferent congestion states are separately controlled by window adjustment strategy.Compared against old mechanism such as TCP and DCCP, the results show the significantimprovement of the network in terms of fairness, packet loss rate and the network throughput.2) MAC layer: Transport layer congestion control mechanisms solve networkcongestion at the cost of throughput. Secondly, the congestion control will significantly lagbehind the occurrence of events while the congestion node farther away from the source.Therefore, the hop-by-hop feedback congestion control mechanism is put forward based onthe analysis with MAC layer. The mechanism uses the number of RTS retransmission andcached queue value as congestion detection, and control network congestion by improving theMAC layer data frame transmission and the backoff mechanism. The simulation results showthe proposed mechanism control the congestion quickly, and improve the channel utilizationand the fairness of the network.3) Routing layer: Multipath routing can balance network load and reduce congestion.However, the existing multipath routing protocol typically uses only the primary path for data transmission, resulting in overloading the main path, node premature death along the path.Hence, an adaptive flow distribution for congestion is proposed to improve the networkquality of service, which based on multipath routing protocol. The relevant factor mode isintroduced in the new strategy to build a low correlation path set. Simultaneously, the newstrategy establishes path congestion information collection, update mechanism. Andeventually source node adaptively selects the lower congestion path to avoid congestion byreal-time path congestion information. The simulation results show the improved strategy caneffectively improve the packet delivery and reduce the end-to-end delay.