Huber cubature particle filter and online state estimation

Abstract

The cubature particle filter, which utilizes the third-degree spherical–radial cubature rule to generate proposal distribution, performs inaccurately when the measurement noises deviate from the Gaussian probability distribution. To improve the performance of cubature particle filter, the update process is recast by making use of the Huber technique which could generate robustness with respect to deviations from the assumed Gaussian noise. Moreover, the general resampling process is replaced by fine variety resampling approach, and as a result the degeneracy phenomenon is effectively reduced. The improved state estimation algorithm is named as Huber cubature particle filter (HCPF). The effectiveness of the proposed algorithm is illustrated via an aircraft actuator model and the experimental results are provided to support the theoretical results.