Unscented Particle Filter Algorithm Based on Divide-and-Conquer Sampling for Target Tracking.

Area of Science:

• Signal Processing
• Control Systems Engineering
• Robotics and Automation

Background:

• Unscented Particle Filter (UPF) faces challenges in maneuvering target tracing due to limited state space coverage.
• Low sample diversity and algorithm redundancy in UPF can degrade tracking performance.
• Maneuvering targets present complex dynamics that challenge traditional filtering methods.

Purpose of the Study:

• To enhance the tracking accuracy and efficiency of the Unscented Particle Filter (UPF) for maneuvering targets.
• To address particle degeneracy and improve sample diversity in UPF algorithms.
• To introduce a novel divide-and-conquer sampling strategy for improved target state estimation.

Main Methods:

• Application of a divide-and-conquer sampling method to the UPF algorithm.
• Decomposition of the target state space for descending dimension processing of maneuvers.
• Separate particle sampling within subspaces to prevent particle degeneracy.

Main Results:

• The proposed Divide-and-Conquer Sampling Unscented Particle Filter (DCS-UPF) demonstrated improved particle diversity.
• DCS-UPF achieved higher tracking accuracy compared to particle swarm and intelligent adaptive filtering algorithms.
• The algorithm showed improved performance in less time, indicating enhanced efficiency.

Conclusions:

• The DCS-UPF effectively improves particle diversity and tracking accuracy for maneuvering targets.
• The divide-and-conquer sampling strategy successfully mitigates UPF limitations in complex dynamic environments.
• This algorithm offers a robust solution for real-time maneuvering target tracing applications.