Evolutionary-Inspired Strategy for Particle Distribution Optimization in Auxiliary Particle Filtering Algorithm Based Indoor Positioning

Particle filter (PF) has been widely used in the target state and position estimations owing to its superiority in tackling the complicated nonlinear problems with arbitrary distributions. As an advanced PF algorithm, Sequential Importance Resampling (SIR) has been widely used for indoor positioning. Since the proposal density of SIR is independent of measurement, the algorithm is vulnerable to outliers. Although the Auxiliary SIR (ASIR) overcomes this problem by performing a two-stage sampling, its positioning accuracy is degraded when the process noise in the filter is large. In order to tackle this problem, an improved ASIR named evolutionary-strategy-integrated ASIR algorithm (EASIR) is proposed in this paper for accuracy improvement in indoor positioning. Tests are carried out for assessing the positioning performance of the proposed algorithm, and positioning accuracy and computation efficiency are considered as the performance metrics in the assessment. Comparing with the SIR and ASIR, the results show that the EASIR achieves better positioning accuracy when the same number of particles are used for data processing and has better robustness when the process noise in the filter is large. Moreover, the computation efficiency of EASIR is generally affordable for real-time applications.