Rao-Blackwellized particle filter for pattern matching indoor localisation

Abstract

Pattern matching localisation based on Received Signal Strength Indication (RSSI) is widely implemented in the Wireless Local Area Network (WLAN). This implementation is commonly with a filtering method to achieve better location estimation accuracy. A Kalman filter (KF) is an optimal filter, if requirements on linear/Gaussian state space model are met. Otherwise, a particle filter (PF) should be used to deal with nonlinear/non-Gaussian state space model. However, in the real situation especially in the localisation field, the state space model may be linear/Gaussian and nonlinear/non-Gaussian. Therefore, there should be a filtering method that can accommodate both linear/Gaussian and nonlinear/non-Gaussian state space model such as Rao Blackwellized particle filter (RBPF). RBPF implemented in the pattern matching localisation system is described and its performance is compared against KF and PF. Those three filtering methods are evaluated in the test bed. To the best of our knowledge, implementing RBPF and performance comparison against KF and PF in the pattern matching indoor localisation in WLAN environment have never been published before.