An Efficient Adaptive Weights Update Scheme for a Gaussian Mixture Filter
LI Cun · 2021
This paper proposes an adaptive weight update scheme of the Gaussian components for the Gaussian mixture filter in the time update stage. This method contributes to obtaining a better approximation of the posterior probability density function, which is constrained by large uncertainty in the measurements or ambiguity in the model. The Gaussian mixture filter is improved through combination with the Cubature Kalman Filter(CKF). The Gaussian components are predicted and updated using a CKF with the results merged and weighted. A series of extensive trails were run to assess the estimation precision offered by various algorithms. The results based on the Unmanned Underwater Vehicle(UUV) lake trial data demonstrate the superiority of the proposed algorithm through better accuracy and stability compared with the conventional navigation algorithms, with are reasonable computational time to meet real-time navigation requirements.