Reversible Jump MCMC for Deghosting in MSPSR Systems

Area of Science:

• Radar Systems Engineering
• Signal Processing
• Statistical Inference

Background:

• Bistatic track association and deghosting are critical challenges in multi-static primary surveillance radar (MSPSR) systems.
• Existing methods often struggle with the complexities of non-cooperative targets and complex signal environments.
• Frequency modulation (FM)-based MSPSR presents unique difficulties for traditional tracking algorithms.

Purpose of the Study:

• To develop and present a novel algorithm for bistatic track association and deghosting in FM-based MSPSR.
• To provide a detailed theoretical framework and algorithmic description, including custom Markov chain Monte Carlo moves.
• To evaluate the proposed algorithm's performance against existing methods using simulated data.

Main Methods:

• A hierarchical Bayesian model is employed, utilizing the Indian buffet process (IBP) as a prior for the association matrix.
• Reversible jump Markov chain Monte Carlo (RJMCMC) is used for inferring the association matrix, incorporating a custom set of sampler moves.
• Simulated data are generated to rigorously test and compare the proposed algorithm.

Main Results:

• The proposed algorithm demonstrates significantly superior performance compared to two alternative methods in simulated bistatic tracking scenarios.
• Analysis of simulated data provides insights into the convergence properties and posterior distributions of the Markov chains generated by the sampler.
• The custom sampler moves are shown to be effective in exploring the complex state space for association inference.

Conclusions:

• The novel IBP-based RJMCMC algorithm offers a robust and effective solution for bistatic track association and deghosting in FM-based MSPSR.
• The method shows promise for enhancing radar tracking capabilities in complex, multi-static environments.
• Further research directions are identified to extend and refine the proposed technique.