Improved Bearings-Only Multi-Target Tracking with GM-PHD Filtering.

Area of Science:

• Signal Processing
• Data Fusion
• Estimation Theory

Background:

• Multi-target tracking with bearings-only measurements presents significant challenges due to nonlinearities and data association ambiguities.
• Existing Gaussian mixture probability hypothesis density (GM-PHD) filters often struggle with accurate posterior intensity approximation and likelihood modeling.
• The extended Kalman filter (EKF) and unscented Kalman filter (UKF) are commonly used but have limitations in nonlinear scenarios.

Purpose of the Study:

• To propose an improved nonlinear Gaussian mixture probability hypothesis density (GM-PHD) filter for enhanced multi-target tracking using bearings-only measurements.
• To address limitations in posterior intensity approximation and likelihood modeling found in conventional GM-PHD filters.
• To introduce a novel target birth model that improves filter performance.

Main Methods:

• Development of the Gaussian mixture measurements-probability hypothesis density (GMM-PHD) filter.
• Approximation of posterior intensity using a Gaussian mixture.
• Modeling the likelihood function with a Gaussian mixture, deviating from single Gaussian assumptions.
• Implementation of a partially uniform target birth model.

Main Results:

• The proposed GMM-PHD filter demonstrates superior performance in multi-target tracking scenarios with bearings-only measurements.
• Simulation results confirm that the GMM-PHD filter outperforms standard GM-PHD filters utilizing EKF and UKF.
• The enhanced likelihood and birth models contribute to improved tracking accuracy and robustness.

Conclusions:

• The GMM-PHD filter offers a significant advancement for bearings-only multi-target tracking.
• The proposed filter provides a more accurate and robust solution compared to existing methods.
• This work contributes to the field of target tracking by improving estimation accuracy in challenging environments.