Adaptive Fifth-Degree Cubature Information Filter for Multi-Sensor Bearings-Only Tracking

Area of Science:

• Signal Processing
• Estimation Theory
• Control Systems

Background:

• Standard Bayesian filtering requires known statistical properties of system noise, which is often unrealistic in target tracking.
• Accurate estimation of system noise covariance is crucial for effective Bayesian filtering performance.

Purpose of the Study:

• To develop a novel adaptive estimation algorithm for multi-sensor bearings-only tracking (BOT) with unknown process noise covariance.
• To enhance the accuracy and stability of target tracking in challenging real-world applications.

Main Methods:

• The proposed adaptive fifth-degree cubature information filter (AFCIF) is based on the fifth-degree cubature Kalman filter and operates within the information filtering framework.
• A sensor selection strategy utilizing observability theory and a recursive process noise covariance estimation based on the covariance matching principle are employed.

Main Results:

• The AFCIF algorithm demonstrates superior estimation accuracy compared to existing methods.
• The proposed filter exhibits enhanced filtering stability, particularly under conditions of unknown process noise covariance.

Conclusions:

• The AFCIF provides a robust solution for multi-sensor bearings-only tracking problems where system noise characteristics are not precisely known.
• The integration of adaptive noise estimation and sensor selection significantly improves tracking performance.