Adaptive Collaborative Gaussian Mixture Probability Hypothesis Density Filter for Multi-Target Tracking.

Area of Science:

• Robotics and Autonomous Systems
• Signal Processing
• Data Fusion

Background:

• Multi-target tracking (MTT) is crucial for systems like autonomous vehicles and surveillance.
• Existing Probability Hypothesis Density (PHD) filters face challenges in computational efficiency and automatic track management.
• Accurate distinction between persistent and newly appearing targets is essential for robust MTT.

Purpose of the Study:

• To propose an adaptive collaborative Gaussian Mixture Probability Hypothesis Density (ACo-GMPHD) filter for enhanced multi-target tracking.
• To enable automatic track extraction and management within the filtering framework.
• To improve both tracking accuracy and computational efficiency compared to existing methods.

Main Methods:

• Adaptive partitioning of measurements into persistent and birth sets based on target evolution.
• Utilizing separate Probability Hypothesis Density (PHD) updates for persistent and birth targets.
• Implementing a novel collaboration mechanism among multiple PHD filters for track extraction.

Main Results:

• The ACo-GMPHD filter demonstrated significant improvements in tracking accuracy.
• Substantial reductions in computational processing requirements were achieved.
• Automatic track extraction was successfully integrated and validated.

Conclusions:

• The proposed ACo-GMPHD filter offers a computationally efficient and accurate solution for multi-target tracking.
• Adaptive measurement partitioning and collaborative PHD mechanisms are effective for robust track management.
• This approach advances the state-of-the-art in autonomous systems and sensor data fusion.