Limited Memory-Based Random-Weighted Kalman Filter

Area of Science:

• Control Engineering
• Signal Processing
• Estimation Theory

Background:

• Kalman filters require precise system noise statistics for optimal state estimation.
• Inaccurate noise statistics in dynamic environments degrade Kalman filter performance.
• Existing methods struggle with unknown or uncertain noise parameters.

Purpose of the Study:

• To develop a robust method for estimating system noise statistics in Kalman filtering.
• To improve the accuracy and stability of state estimation under uncertain noise conditions.
• To combine random weighting with limited memory techniques for adaptive noise estimation.

Main Methods:

• Utilizing the random weighting concept to establish theories for noise statistics estimation.
• Applying the limited memory technique to focus on recent historical data for estimation.
• Integrating estimated process and measurement noise statistics back into the Kalman filter.
• Adaptively adjusting the weights of system noise statistics within a limited memory.

Main Results:

• The proposed method accurately estimates both process and measurement noise statistics.
• Improved Kalman filtering accuracy and stability demonstrated through simulations and experiments.
• The method effectively suppresses interference from system noise on state estimation.
• Overcame limitations of traditional limited memory filters.

Conclusions:

• The novel approach enhances Kalman filter accuracy by adaptively estimating system noise statistics.
• Combining random weighting and limited memory provides a robust solution for uncertain noise environments.
• This technique offers improved system state estimation in practical dynamic systems.