论文信息 - Comparison of Particle Method and Finite Difference Nonlinear Filters for Low SNR Target Tracking

Comparison of Particle Method and Finite Difference Nonlinear Filters for Low SNR Target Tracking

This paper presents the formulation of a low SNR target tracking challenge problem that we have designed to explore the numerical and estimation performance trades of nonlinear filtering methods. We have implemented a particle filter (PF) and alternating direction implicit-based (ADI) finite difference method filter. Our initial result from comparing these methods in terms of their RMS position error is that PF betters ADI by about 60% at 2.5dB SNR, the point of maximum error reduction. More significantly, PF also appears to be much more robust against track loss at these low SNR levels. This robustness derives from the inherent adaptivity of the particle method.

Stanton Musick | John H. Greenewald | J. Greenewald | S. Musick

[1] A. Jazwinski. Stochastic Processes and Filtering Theory , 1970 .

[2] N. N. Yanenko. Application of the Method of Fractional Steps to Boundary Value Problems for Laplace’s and Poisson’s Equations , 1971 .

[3] Simon J. Godsill,et al. On sequential simulation-based methods for Bayesian filtering , 1998 .

[4] Harold W. Sorenson,et al. On the development of practical nonlinear filters , 1974, Inf. Sci..

[5] Keith D. Kastella,et al. Nonlinear filtering for tracking low-elevation targets in multipath , 1998, Defense, Security, and Sensing.

[6] B. Rozovskii,et al. Nonlinear Filtering Revisited: A Spectral Approach , 1997 .

[7] J. Strikwerda. Finite Difference Schemes and Partial Differential Equations , 1989 .

[8] D. Rubin. Using the SIR algorithm to simulate posterior distributions , 1988 .

[9] N. Gordon,et al. Novel approach to nonlinear/non-Gaussian Bayesian state estimation , 1993 .