Belangrijkste concepten
Efficient point-mass filter design for terrain-aided navigation improves accuracy and reduces computational complexity.
Samenvatting
The content discusses the design of an efficient point-mass filter for state estimation in stochastic models with linear dynamics. It introduces a novel efficient PMF (ePMF) estimator that unifies continuous and discrete approaches, reducing time complexity significantly without compromising accuracy. The paper also provides MATLAB® code for the ePMF implementation. The article covers various models, grid movements, and algorithms to enhance the efficiency of the point-mass filter.
I. Introduction
- State estimation of stochastic dynamic systems from noisy measurements.
- Bayesian recursive relations computing probability density functions.
- Approximate filtering methods divided into global and local filters.
II. Terrain Aided Navigation and State Estimation
- Utilization of terrain altitude measurements for navigation.
- Presentation of continuous and discrete dynamics models.
III. Point-Mass Filter
- Approximation of conditional PDF by piece-wise constant point-mass density.
- Basic algorithm steps for the Point-Mass Filter.
IV. Efficient Point-Mass Prediction
- Transformation of time-update to convolution for computational efficiency.
- Use of Fast Fourier Transform (FFT) and Fast Sine Transform (FST) based solutions.
V. Efficient Point-Mass Filter Design
- Grid movement compensation to address degeneracy issues.
- Design considerations for continuous ePMF with non-diagonal noise.
VI. Enclosed MATLAB® Codes
- Availability of MATLAB® codes on IDM research team's website.
VII. Numerical Results
Two-Dimensional Estimation Results:
- Comparison between PMF, FFT-based PMF, FST-based PMF, and PF bootstrap.
Four-Dimensional Estimation Results:
- Comparison between FFT-based PMF and PF bootstrap in four dimensions.
VIII. Concluding Remarks
- Proposal of an efficient point-mass filter design for terrain-aided navigation.
Statistieken
The proposed ePMF can lead to a time complexity reduction exceeding 99.9% without compromising accuracy.
Citaten
"The proposed ePMF can now compete with PF in both accuracy and computational complexity."