Variational Bayesian Algorithms for Maneuvering Target Tracking with Nonlinear Measurements in Sensor Networks.

Area of Science:

• Signal Processing
• Estimation Theory
• Networked Systems

Background:

• Variational Bayesian methods are crucial for nonlinear estimation problems, but their performance is sensitive to linear approximations and system nonlinearity.
• Maneuvering target tracking with Doppler measurements in networked sensor systems presents significant estimation challenges.

Purpose of the Study:

• To propose a distributed fusion variational Bayesian Kalman filter for networked maneuvering target tracking.
• To evaluate the performance of the proposed filter using evidence lower bound and posterior Cramér-Rao lower bound.

Main Methods:

• Implementing a variational Bayesian Kalman filter using natural gradient and simultaneous perturbation stochastic methods.
• Developing a distributed fusion approach for networked sensor data under single-hop constraints.
• Analyzing theoretical bounds including evidence lower bound and posterior Cramér-Rao lower bound.

Main Results:

• The proposed distributed fusion filter demonstrates improved performance in maneuvering target tracking.
• Simulations show the distributed method outperforms centralized fusion in terms of posterior Cramér-Rao lower bounds and root-mean-squared errors.
• The 3σ bound analysis further validates the enhanced accuracy of the proposed approach.

Conclusions:

• The distributed fusion variational Bayesian Kalman filter is an effective solution for networked maneuvering target tracking.
• The proposed methods offer a robust and accurate estimation framework for complex, nonlinear, and networked scenarios.
• This work advances the state-of-the-art in distributed estimation for sensor networks.