Constraint Kalman filter for indoor bluetooth localization

This paper studies sequential estimation of indoor localization based on fingerprints of received signal strength indicators (RSSI). Due to the lack of an analytic formula for the fingerprinting measurements, the Kalman filter can not be directly applied. By introducing a hidden variable to represent the unknown positioning coordinate, a state model is formulated and a constrained Kalman filter (CKF) is then derived within the Bayesian framework. The update of the state incorporates the prior information of the motion model and the statistical property of the hidden variable estimated from the RSSI measurements. The positioning accuracy of the proposed CKF method is evaluated in indoor field tests by a self-developed Bluetooth fingerprint positioning system. The conducted field tests demonstrate the effectiveness of the method in providing an accurate indoor positioning solution.