Optimal weighted K‐nearest neighbour algorithm for wireless sensor network fingerprint localisation in noisy environment. Issue 10 (1st May 2018)

• Record Type: Journal Article
• Authors: Fang, Xuming; Jiang, Zonghua; Nan, Lei; Chen, Lijun
• Is Part Of: IET communications. Volume 12:Issue 10(2018)
• Page Start: 1171
• Page End: 1177
• Publication Date: 2018-05-01
• Subjects: wireless sensor networks -- sensor placement -- nearest neighbour methods -- adaptive filters -- Kalman filters -- RSSI; optimal weighted K‐nearest neighbour algorithm -- wireless sensor network -- fingerprint localisation algorithm -- noisy environment -- WKNN algorithm -- fingerprint positioning -- node position estimation -- optimised node location estimate -- OWKNN algorithm -- WSN -- adaptive Kalman filter -- AKF -- memetic algorithm -- measurement noise reduction -- received signal strength indication -- RSSI -- calibration point weight optimization -- target node position estimation -- calibrated radio map; Telecommunication systems -- Periodicals; Speech processing systems -- Periodicals; 621.38205
• DOI: 10.1049/iet-com.2017.0515
• Languages: English
• ISSNs: 1751-8628

Abstract

Abstract : The weighted K‐nearest neighbour (WKNN) algorithm is widely applied to fingerprint positioning. However, the node position estimated by the WKNN algorithm is not optimal in a noisy environment. To obtain the optimised node location estimate, the authors propose an optimal WKNN (OWKNN) algorithm for wireless sensor network (WSN) fingerprint localisation in a noisy environment. The proposed OWKNN algorithm is composed of an adaptive Kalman filter (AKF) and a memetic algorithm (MA). First, the AKF is utilised to reduce the measurement noise of the received signal strength indication (RSSI) between the nodes in the WSN. Then, the MA is employed to optimise the calibration point weight for estimating the position of a target node in the WSN according to the filtered RSSI and a calibrated radio map. Finally, an optimal node location estimate is achieved based on the optimised weight. The extensive experimental results reveal that the localisation accuracy of the proposed algorithm is at least ∼50% higher than those of the state‐of‐the‐art fingerprint localisation algorithms regardless of the placement of the target node, number of beacon nodes, and size of the calibration cell.