Weighted Optimization-Based Distributed Kalman Filter for Nonlinear Target Tracking in Collaborative Sensor Networks.

Area of Science:

• Signal Processing
• Control Systems
• Networked Systems

Background:

• Nonlinear target tracking in collaborative sensor networks is vital.
• Accurate identification of nonlinearity and coupling is essential.
• Adaptive Kalman filtering (KF) methods struggle with rapidly changing nonlinear systems due to large data window requirements.

Purpose of the Study:

• To propose a novel weighted optimization-based distributed KF algorithm (WODKF).
• To address the limitations of traditional KF methods in nonlinear, rapidly changing environments.
• To enhance model noise covariance estimation for improved tracking accuracy.

Main Methods:

• The WODKF algorithm expands data size using measurements and estimates from connected sensors.
• A new cost function, a weighted sum of state bias and oscillation, is introduced for model noise covariance estimation.
• Polynomial fitting and an exhaustive method are used to compute the optimal model noise covariance.
• Sensor selection is incorporated to reduce computational load and increase scalability.
• Local probability data association is employed for multitarget tracking.

Main Results:

• The WODKF algorithm effectively estimates model noise covariance by leveraging distributed sensor data.
• Simulations demonstrate the feasibility and superiority of WODKF compared to other filtering algorithms.
• The algorithm shows robust performance in tracking random signals and multiple nonlinear targets.

Conclusions:

• The proposed WODKF algorithm offers a practical and superior solution for nonlinear target tracking in collaborative sensor networks.
• The method overcomes the limitations of traditional adaptive KF by utilizing distributed data effectively.
• WODKF enhances tracking accuracy and network scalability for a wide range of systems.