A Robust SMC-PHD Filter for Multi-Target Tracking with Unknown Heavy-Tailed Measurement Noise

Area of Science:

• Signal Processing
• Statistical Inference
• Robotics

Background:

• The Sequential Monte Carlo Probability Hypothesis Density (SMC-PHD) filter is widely used for multi-target tracking.
• Performance degradation occurs due to outliers and unknown heavy-tailed measurement noise.
• Existing methods struggle with non-Gaussian noise models.

Purpose of the Study:

• To propose a robust SMC-PHD (RSMC-PHD) filter for improved multi-target tracking.
• To address performance degradation caused by unknown heavy-tailed measurement noise.
• To enhance tracking accuracy and reliability in cluttered environments.

Main Methods:

• Introduction of the Student-t distribution to model unknown heavy-tailed measurement noise.
• Modeling degrees of freedom (DOF) and scale matrix using Gamma and inverse Wishart distributions, respectively.
• Application of Variational Bayesian (VB) technique for parameter inference and derivation of the RSMC-PHD filter recursion.
• Implementation of a particle weight modification strategy to prevent target number overestimation.

Main Results:

• The proposed RSMC-PHD filter demonstrates effectiveness in the presence of unknown heavy-tailed measurement noise.
• Simulations confirm significant performance improvements compared to standard SMC-PHD filters.
• The filter robustly estimates target states and numbers under noisy conditions.

Conclusions:

• The developed RSMC-PHD filter offers a robust solution for multi-target tracking with heavy-tailed noise.
• The integration of Student-t distribution and VB inference enhances tracking accuracy.
• This approach provides a practical and effective method for real-world tracking scenarios.