Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Comparative analysis of ellipsoidal methods for distributed data fusion
Bakr M., Lee S.  IMCOM 2018 (Proceedings of the 12th International Conference on Ubiquitous Information Management and Communication, Langkawi, Malaysia, Jan 5-7, 2018)1-8.2018.Type:Proceedings
Date Reviewed: Jul 11 2018

In multisensor systems, the fusion of information from multiple sensors can be done globally or locally in distributed “fusion centers.” The distributed approach has advantages in reliability, fault tolerance, and efficiency, but can be difficult because the specifics of the correlation between data from different sensors are generally unknown (or awkward when known). The authors address the problem of distributed data fusion, specifically the combination of multiple sensor estimates with different variances.

Assuming that noise is normally distributed and that sensor outputs are somewhat correlated, the distribution of the sensed values is multivariate normal, but the cross-correlation may not be precisely known. Fusing the separate sensor data could be done by maximum likelihood estimation, but this assumes knowledge of the cross-correlations. Two types of methods have been proposed to deal with imprecise knowledge of the cross-correlations: data decorrelation and ellipsoidal methods, of which the latter is preferred due to its efficiency and accuracy. This paper compares several of the ellipsoidal methods.

Essentially, ellipsoidal methods for data fusion consider a set of ellipses, each of which forms a bound on the individual sensor estimates. The combined sensor estimates are then represented by a new ellipse computed in various ways from the individual sensor estimate ellipses. The covariance intersection method results in a new ellipse that bounds the intersection of the individual ellipses; it is well suited for two data sources, but is a weak bound (and computationally expensive) for more sources. The largest ellipsoid method produces an ellipse that is itself bounded by the intersection area; it is efficient, but the mean estimate is biased. The internal ellipsoid is based on a heuristic-guided fit to the multiple ellipses and provides a better mean estimate, whereas the ellipsoidal intersection method yields good results but is only defined for the two-sensor case.

As applications of distributed sensor fusion become more demanding, approaches like this will be valuable in achieving the promise of the technology. The paper provides well-presented mathematical descriptions of the methods and of their results on sample data. The figures are helpful in describing the method. For those engaged in this specific area of sensor fusion, the paper is an important resource.

Reviewer:  Creed Jones Review #: CR146142 (1809-0503)
Bookmark and Share
 
Distributed Systems (C.2.4 )
 
 
Robotics (I.2.9 )
 
 
Scene Analysis (I.4.8 )
 
 
General (E.0 )
 
 
Probability And Statistics (G.3 )
 
Would you recommend this review?
yes
no
Other reviews under "Distributed Systems": Date
The evolution of a distributed processing network
Franz L., Sen A., Rakes T. Information and Management 7(5): 263-272, 1984. Type: Article
Jul 1 1985
A geographically distributed multi-microprocessor system
Angioletti W., D’Hondt T., Tiberghien J.  Concurrent languages in distributed systems: hardware supported implementation (, Bristol, UK,871985. Type: Proceedings
Oct 1 1985
A fault tolerant LAN with integrated storage, as part of a distributed computing system
Boogaard H., Bruins T., Vree W., Reijns G.  Concurrent languages in distributed systems: hardware supported implementation (, Bristol, UK,1001985. Type: Proceedings
Aug 1 1985
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy