Gauge-invariant registration in networks

Stephen D. Howard, Douglas Cochran, William Moran

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Integration and exploitation of information collection across a distributed network of assets usually requires the establishment and maintenance of registration of coordinates across the nodes of the network. Here 'registration' covers a range of possibilities, including clock synchronization and registration of frames of reference. The registration problem is posed in terms of network represented by a graph Γ with vertices corresponding to the sensors. Attached to the edges of the graph are noisy measurements of the 'difference' between the two coordinate systems. This 'difference' is expressed in terms of a member of a Lie group of coordinate transformations. Effectively, the registration problem is specified in terms of a connection on the edges, and becomes one of estimating a gauge transformation to align the coordinate systems across the network. The key descriptor of the difficulty of the estimation problem, the Fisher information, can be simply expressed in terms of the geometry of the situation and provides a link between the homological chains and cochains for the graph.

Original languageEnglish (US)
Title of host publication2015 18th International Conference on Information Fusion, Fusion 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1526-1532
Number of pages7
ISBN (Print)9780982443866
StatePublished - Sep 14 2015
Event18th International Conference on Information Fusion, Fusion 2015 - Washington, United States
Duration: Jul 6 2015Jul 9 2015

Other

Other18th International Conference on Information Fusion, Fusion 2015
Country/TerritoryUnited States
CityWashington
Period7/6/157/9/15

ASJC Scopus subject areas

  • Information Systems
  • Signal Processing
  • Computer Networks and Communications

Fingerprint

Dive into the research topics of 'Gauge-invariant registration in networks'. Together they form a unique fingerprint.

Cite this