Centralized Fusion Approach to the Estimation Problem with Multi-Packet Processing under Uncertainty in Outputs and...

Area of Science:

• Control Systems Engineering
• Signal Processing
• Networked Systems

Background:

• Multi-sensor network systems face challenges with data uncertainties and unreliable communication channels.
• Packet dropouts and random delays in data transmission degrade estimation performance.
• Accurate state estimation is crucial for effective decision-making in networked systems.

Purpose of the Study:

• To develop a least-squares linear centralized estimation method for multi-sensor networks.
• To account for uncertainties in measurements modeled by random parameter matrices.
• To handle random one-step delays and packet dropouts during data transmission.

Main Methods:

• Utilizing augmented observation vectors to combine sensor measurements.
• Employing an innovation approach to derive recursive estimation algorithms.
• Designing estimators that compensate for missing data using predicted outputs.
• Introducing Bernoulli random variables to model observation uncertainties.

Main Results:

• Centralized fusion estimators (predictors, filters, smoothers) are obtained recursively.
• The proposed method does not require the signal evolution model.
• A numerical example demonstrates the model's applicability to uncertain systems with state-dependent noise.
• Estimation accuracy is shown to be influenced by sensor uncertainties and transmission failures.

Conclusions:

• The developed method provides robust centralized estimation in multi-sensor networks with uncertain measurements and unreliable transmissions.
• Recursive algorithms offer an efficient way to compute estimators without needing the signal model.
• The findings highlight the impact of network unreliability on estimation performance, offering insights for system design.