Gaussian Mixture Cardinalized Probability Hypothesis Density(GM-CPHD): A Distributed Filter Based on the Intersection...

Area of Science:

• Signal Processing
• Estimation Theory
• Multisensor Data Fusion

Background:

• Accurate sensor signal processing is crucial for reliable system performance.
• Distributed filtering methods face challenges from local filtering and time-varying noise.
• Existing algorithms may struggle with computational complexity and nonlinear systems.

Purpose of the Study:

• To design a distributed GM-CPHD filter to mitigate noise and improve sensor signal accuracy.
• To develop an efficient data fusion algorithm that reduces computational load and processing time.
• To enhance the generalization capability and reduce nonlinear complexity in multisensor systems.

Main Methods:

• Utilized the Gaussian mixture-cardinalized probability hypothesis density (GM-CPHD) filter for subsystem estimation.
• Implemented an inverse covariance cross-fusion algorithm for signal merging and convex optimization.
• Integrated the GM-CPHD filter into the conventional inverse covariance intersection (ICI) structure, creating the PICI-GM-CPHD algorithm.

Main Results:

• The PICI-GM-CPHD algorithm demonstrated a smaller OSPA error compared to mainstream algorithms in simulations.
• The improved algorithm significantly enhanced signal processing accuracy for both linear and nonlinear signals.
• Reduced computational burden and data fusion time were observed, indicating improved efficiency.

Conclusions:

• The developed PICI-GM-CPHD algorithm offers a practical and advanced solution for multisensor data processing.
• The algorithm effectively attenuates noise and improves the accuracy of sensor signal estimation.
• This approach presents a significant advancement in handling complex, noisy, and time-varying data in distributed systems.