Adaptive Unscented Kalman Filter for Target Tracking with Unknown Time-Varying Noise Covariance

Area of Science:

• Signal Processing
• Control Systems Engineering

Background:

• The unscented Kalman filter (UKF) is a standard algorithm for nonlinear target tracking.
• Standard UKF performance degrades with noise covariance mismatches, and existing adaptive UKF algorithms have limitations.

Purpose of the Study:

• To propose a new adaptive UKF scheme to address time-varying noise covariance problems in target tracking.
• To enhance the stability and accuracy of target tracking under uncertain noise conditions.

Main Methods:

• Derivation and proof of the cross-correlation between innovation and residual sequences.
• Real-time estimation of process noise covariance using a linear matrix equation derived from innovation and residual sequences.
• Estimation of measurement noise covariance using an improved measurement-based adaptive Kalman filtering algorithm with redundant measurements.

Main Results:

• The proposed adaptive UKF demonstrates superior performance compared to the standard UKF and current adaptive UKF algorithms.
• The algorithm achieves improved tracking accuracy and stability under time-varying noise covariances.
• The measurement noise covariance estimation is immune to state estimation errors.

Conclusions:

• The developed adaptive UKF scheme effectively resolves time-varying noise covariance issues in target tracking.
• This advancement offers a more robust and accurate solution for nonlinear filtering problems.
• The proposed method enhances the reliability of target tracking systems in dynamic environments.