2014/01/13 by Pete Bunch, Bunch, Pete, Simon Godsill +1
Computer Science · Engineering · Environmental Science · Mathematics · #Computation (stat.CO) #FOS: Computer and information sciences #Hydrological Forecasting Using AI #Target Tracking and Data Fusion in Sensor Networks #Traffic Prediction and Management Techniques #stat.CO
paper · pdf · doi:10.48550/arxiv.1401.2791
Superseded by arXiv:1406.3183 Changes include a substantially revised presentation and several errors corrected
openalex publication_date 2014/01/13 · arxiv created 2014/08/14 · arxiv updated 2014/08/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The crucial step in designing a particle filter for a particular application is the choice of importance density. The optimal scheme is to use the conditional posterior density of the state, but this cannot be sampled or calculated analytically in most case. In practice, approximations of this density are used, particularly Gaussian densities based on linearisation or the unscented transform. For many highly nonlinear or non-Gaussian models, these approximations can be poor, leading to degeneracy of the particle approximation or even the filter "losing track" completely. In this paper, we develop a new mechanism for approximating the optimal importance density, which we call the progressive proposal method. This works by introducing the observation progressively and performing a series of state updates, each using a local Gaussian approximation to the optimal importance density. A number of refinements and extensions to the basic algorithm are also introduced. Simulations are used to demonstrate an improvement in performance over simpler particle filters on a number of applications.