RS Algorithms


RS algorithms that significantly outperform conventional approaches:


  • Single-target tracking in arbitrary clutter and detection profiles: A Bernoulli filter that significantly outperforms traditional single-target filters such as PDA. 
    B. Ristic, B.-T. Vo, B.-N. Vo, and A. Farina "A Tutorial on Bernoulli Filters: Theory, Implementation and Applications," IEEE Trans. Signal Processing, Vol. 61, No. 13, pp. 3406 - 3430, 2013

  • Track-Before-Detect (TBD) in pixelized images: A provably Bayes-optimal multi-Bernoulli filter that significantly outperforms the previously best TBD algorithm, histogram-PMHT.
    Chapter 20 of Mahler, Advances in Statistical Multisource-Multitarget Information Fusion, 2014 

  • TBD for superpositional sensors: A CPHD filter for passive-acoustic, RF-tomography, and other superpositional sensors that significantly outperforms conventional MCMC methods.
    - Chapter 19 of Mahler, Advances in Statistical Multisource-Multitarget Information Fusion, 2014

  • Simultaneous localization and mapping (SLAM):  A robotics PHD-SLAM algorithm that, in heavy clutter, significantly outperforms conventional approaches such as MH-FastSLAM.
    - Mullane et al., Random Finite Sets in Robotic Map Building and SLAM, Springer (2011)

  • Multitarget tracking in heavy clutter: “Classical” CPHD filters that can track in very heavy clutter backgrounds that challenge traditional combinatorial algorithms
    - Chapters 8, 9 of Mahler, Advances in Statistical Multisource-Multitarget Information Fusion, 2014


Recent RS developments worth keeping an eye on:

  • First provably Bayes-optimal and practical multitarget tracker: The first exact closed-form solution to the multitarget Bayes filter.  Already being used in autonomous automobiles.  Real-time operation on a cellphone.
    Chapter 15 of Mahler, Advances in Statistical Multisource-Multitarget Information Fusion, 2014    
    - B.-T. Vo, and B.-N. Vo, "Labeled Random Finite Sets and Multi-Object Conjugate Priors," IEEE Trans. Signal Processing, Vol. 61, No. 13, pp. 3460-3475, 2013.  
    - B.-N. Vo, B.-T. Vo, and D. Phung, "Labeled Random Finite Sets and the Bayes Multi-Target Tracking Filter," IEEE Trans. Signal Processing, Vol. 62, No. 24, pp. 6554-6567, 2014   
    - H. Hoang, B.-T. Vo, and B-.N. Vo "Stochastic Sampling Based Implementations of Generalized Labeled Multi-Bernoulli Filters," arXiv
    - S. Reuter, B.-T. Vo, B.-N. Vo, and K. Dietmayer, "The labelled multi-Bernoulli filter," IEEE Trans. Signal Processing, Vol. 62, No. 12, pp. 3246-3260, 2014.

  • First general and theoretically rigorous track-to-track fusion: An exponential-mixture consensus-fusion CPHD filter that is immune to unknown double-counting in networks.
    - Battistelli et al., “Consensus CPHD filter for distributed multitarget tracking,” IEEE J. Sel. Topics Sign. Proc., 7(3): 508-520, 2013
    - Battistelli et al., “Distributed fusion of multitarget densities and and consensus CPHD filters,” Proc. SPIE, Vol. 9474, 2015

  • "Background-agnostic” RS filters: RS filters that operate in unknown, dynamically changing clutter and detection backgrounds  
    - Chapters 17, 18 of Mahler, Advances in Statistical Multisource-Multitarget Information Fusion, 2014 . 
    - R. Mahler, "CPHD filters with unknown quadratic clutter generators," Proc. SPIE, Vol. 9474, 2015.

  • Provably Bayes-optimal fusion of “hard” and “soft” data: Bayes filters that optimally fuse conventional measurements with attributes, features, rules, and natural language. 
    - Chapter 22 of Mahler, Advances in Statistical Multisource-Multitarget Information Fusion, 2014