Localization in Wireless Sensor Networks via Support Vector Regression

• Published in: 2009 3rd International Conference on Genetic and Evolutionary Computing pp. 549 - 552
• Main Authors: Wang, Yong, Xu, Xiaobu, Tao, Xiaoling
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2009
• Subjects: Computer networks; Event detection; Genetics; localization; LQI; Query processing; Radio frequency; RF signals; RSSI; Signal to noise ratio; Sun; SVR; Wireless sensor networks; Working environment noise; WSN
• ISBN: 9781424452453; 1424452457; 9780769538990; 0769538991
• DOI: 10.1109/WGEC.2009.79

For the problems of traditional RSSI localization inaccurate and modeling difficult in WSN, this paper puts forward a support vector regression (SVR) learning algorithm based on RSSI and LQI. By training the samples with RSSI and LQI values as input while coordinates as output, we get the localization model. It differs from other RF-based algorithm in that it can estimate node locations directly according to the RF signals, more importantly it needs only one anchor node at least. Benefited from the good generalization ability of SVR, the algorithm can reach about 1-m location accuracy in complex environment, especially suitable for indoor localization. This paper aims at providing a low-cost, high-accuracy RF-based localization technique.