Conservative Quantization of Covariance Matrices with Applications to Decentralized Information Fusion

Area of Science:

• Networked systems
• Information fusion
• Data compression

Background:

• Information fusion in networked systems faces bandwidth limitations for transmitting high-dimensional data.
• Compressing estimates and covariance matrices can compromise data properties and fusion reliability.

Purpose of the Study:

• To present novel quantization methods for estimates and covariance matrices in networked information fusion.
• To demonstrate the application of these methods with optimal fusion formulas and covariance intersection.

Main Methods:

• Development of quantization techniques for both estimates and covariance matrices.
• Integration of quantization methods with established fusion algorithms like optimal fusion and covariance intersection.
• Performance evaluation through simulations under various data reduction scenarios.

Main Results:

• Proposed quantization methods significantly decrease bandwidth requirements for data transmission.
• Unbiasedness and conservativeness of fusion methods are preserved.
• Effectiveness demonstrated even with substantial data reduction.

Conclusions:

• Quantization offers an effective solution for bandwidth constraints in networked information fusion.
• The presented methods enable reliable fusion without sacrificing essential data properties.
• Simulations confirm the practical utility and efficiency of the proposed approach.