Optimal Distributed Finite-Time Fusion Method for Multi-Sensor Networks under Dynamic Communication Weight.

Area of Science:

• Distributed Systems
• Sensor Networks
• Estimation Theory

Background:

• Distributed state estimation in sensor networks faces challenges with fusion errors due to incomplete node information.
• Existing methods may struggle with convergence and accuracy in dynamic network environments.

Purpose of the Study:

• To develop a novel optimal distributed finite-time fusion filtering method for sensor networks.
• To address fusion errors and improve state estimation accuracy using dynamic communication weights.
• To achieve fast finite-time convergence for global information aggregation.

Main Methods:

• Constructed a local filtering algorithm architecture for fusion error convergence within limited iterations.
• Determined the maximum number of iterations based on the communication topology graph diameter.
• Employed matrix weight fusion for optimal estimation (minimum variance) of local filtering results.
• Introduced Generalized Information Quality (GIQ) to derive relative communication weights, integrated into the fusion algorithm.

Main Results:

• The proposed method achieves fusion error convergence within a finite number of iterations.
• Optimal estimation with minimum variance is attained through matrix weight fusion.
• Dynamic communication weights, based on GIQ and local bias, enhance fusion accuracy.
• Numerical simulations and experimental tests validated the algorithm's effectiveness and feasibility.

Conclusions:

• The developed optimal distributed finite-time fusion filtering method effectively minimizes errors in sensor networks.
• Dynamic communication weights significantly improve the accuracy of distributed state estimation.
• The algorithm demonstrates fast finite-time convergence and practical applicability.