Noise-aware fingerprint localization algorithm for wireless sensor network based on adaptive fingerprint Kalman filter

Abstract

The indoor localization technique is one of the key technologies in the field of wireless sensor network (WSN) research. Fingerprint localization that uses the received signal strength indication (RSSI) is a type of indoor positioning technology that can be applied to a noisy WSN environment. This work proposes a noise-aware fingerprint localization algorithm for WSNs based on an innovative adaptive fingerprint Kalman filter (AFKF), which is effective at refining the computational results of the state-of-the-art fingerprint positioning algorithm in noisy environments. This novel AFKF is composed of a Kalman filter (KF), a noise covariance estimator (NCE) and a fingerprint Kalman filter (FKF). First, the KF filters the RSSI with the measurement noise, and then, the NCE is aware of the noise covariance of the filtered RSSI. Finally, the FKF refines the node position that is estimated by the existing fingerprint localization algorithm according to the filtered RSSI and the perceived noise covariance. Our proposed algorithm not only overcomes the limitation of the current range-based localization algorithm but also solves the problem of the present fingerprint localization algorithm; in other words, it can be applied in a situation in which an accurate RSSI-distance model cannot be established and is applicable to a scenario that has unknown or time-varying noise. The results of practical experiments and numerical simulations show that regardless of how the target nodes are placed or how many beacon nodes there are as well as whether the measurement noise is strong or weak or whether the calibration cell is large or small, the proposed algorithm improves the accuracy of the widely applied fingerprint positioning algorithms by at least 50%. These algorithms include the nearest neighbor algorithm (NN), the K-nearest neighbor algorithm (KNN), the weighted K-nearest neighbor algorithm (WKNN), and their refinement algorithms, namely, the position Kalman filter (PKF) and the FKF.