Study On Link Quality Estimation Mechanism For Wireless Sensor Networks Based On Least Square Support Vector Classification

As WSN (Wireless Sensor Network) rapid development of technology, WSN iswidely used in military, smart home, health, space exploration, urban transportationand other fields. In wireless sensor networks, wireless link quality estimation of theperformance of the network has an important role in the foundation, and can affect thedesign of the upper layer routing protocols. Because the underlying WSN is a wirelesscommunication link shared channel, low-power wireless sensor transceiver betweencommunication networks tend to be more sensitive to variations of the networkchannels, electromagnetic waves to transmit data, there will be losses, may cause lossof data transmission. The current communication link is designed for accurateestimation of the upper layer routing protocol is critical, therefore, for an accurateestimation of the current communication link is a network WSN achieve efficient,reliable communication basis.This paper first reviews the basic concepts of the wireless sensor network link qualityestimation and gives the basic principles of estimation and the specific analysis.Secondly, detailing the least squares support vector machine modeling estimationmodel (LQE-LSSVC) process, including analysis of the training sample data (CCI, LQI,PRR), on support vector machine classification algorithm (ECC) and select, kernelfunction and its parameters OK, and then pass the actual experiment to verify thecorrelation of selected metrics and establish a link quality estimation based on leastsquares support vector machine model, and finally through the experiment, the modelis verified accuracy.Experiments show LQE-LSSVC sensor networks based on link quality estimationmodel was able to accurately larger upper routing protocols to provide technicalsupport, and compared to the PRR link quality estimation mechanism based onstatistics, the model can be used relatively small the probe packets to get a moreaccurate estimation of the value of the link quality, to some extent, avoiding the largenumber of probe packets sent over to bring the cost of energy and prolong the networklifetime. Meanwhile, compared to curve fitting method based on estimation modeling,the model can be accurately assessed for different network environments.