Adaptive Square-Root Unscented Particle Filtering Algorithm for Dynamic Navigation.

Area of Science:

• Control Engineering
• Signal Processing
• Data Fusion

Background:

• Nonlinear filtering is crucial for state estimation in dynamic systems.
• Particle filters can suffer from particle degeneracy and instability due to model noise.
• Existing methods struggle with disturbances from kinematic model noise and abnormal observations.

Purpose of the Study:

• To develop a robust nonlinear filtering algorithm that addresses particle degeneracy and data instability.
• To improve the performance of integrated navigation systems by reducing the impact of noise.
• To enhance the accuracy and reliability of state estimation in challenging environments.

Main Methods:

• Combines adaptive filtering and square-root filtering within an unscented particle filter framework.
• Introduces an adaptive factor adjustment based on predicted residuals to counter abnormal observations and model noise.
• Employs Cholesky factorization to stabilize covariance matrices of predicted state and observation vectors.

Main Results:

• The proposed adaptive square-root unscented particle filtering algorithm demonstrates enhanced stability and noise suppression.
• Experiments show significant improvements in performance for integrated navigation systems.
• The algorithm effectively prevents particle degeneracy and mitigates the effects of abnormal observations.

Conclusions:

• The developed algorithm offers a robust solution for nonlinear filtering challenges.
• It provides superior performance in integrated navigation systems compared to existing methods.
• The combination of adaptive and square-root techniques leads to more reliable state estimation.