Fingerprint localisation algorithm for noisy wireless sensor network based on multi‐objective evolutionary model. Issue 8 (19th May 2017)

• Record Type: Journal Article
• Authors: Fang, Xuming; Nan, Lei; Jiang, Zonghua; Chen, Lijun
• Is Part Of: IET communications. Volume 11:Issue 8(2017)
• Page Start: 1297
• Page End: 1304
• Publication Date: 2017-05-19
• Subjects: indoor navigation -- wireless sensor networks -- evolutionary computation -- RSSI -- calibration -- Kalman filters -- covariance analysis -- search problems; fingerprint localisation algorithm -- noisy wireless sensor network -- multiobjective evolutionary model -- received signal strength indication -- indoor positioning -- measurement noise -- RSSI -- calibration point -- fingerprint positioning algorithm -- noisy WSN -- Kalman filter -- noise covariance estimator -- optimised weight search -- perceived noise covariance; Telecommunication systems -- Periodicals; Speech processing systems -- Periodicals; 621.38205
• DOI: 10.1049/iet-com.2016.1229
• Languages: English
• ISSNs: 1751-8628

Abstract

Abstract : Fingerprint localisation technology using received signal strength indication (RSSI) has become one of the hot spots in the research field of indoor positioning based on wireless sensor networks (WSNs). Due to the presence of the measurement noise of the RSSI, the weight of the calibration point in the current fingerprint positioning algorithm is not optimised. The authors propose a fingerprint localisation algorithm for noisy WSNs based on an innovative multi‐objective evolutionary model. The proposed algorithm at first employs the Kalman filter to filter the abnormal RSSI value, and then utilises a noise covariance estimator to perceive the noise covariance of the RSSI. Finally, the multi‐objective evolutionary model is used to search for the optimised weight of the calibration point via the filtered RSSI and the perceived noise covariance. That the novel evolutionary model can find the best fingerprint estimate with the optimised weight has been proven theoretically in this work. Extensive experimental results on an off‐the‐shelf WSN testbed show that the authors' proposed algorithm improves the accuracy of the state‐of‐the‐art fingerprint positioning algorithm by at least 50% regardless of the placement of the target node, the number of beacon nodes, the size of the calibration cell, and the number of nearest neighbours.