David Broman

Thesis proposal Kevin LeBlanc:

Tentative abstract!


Geometric spaces for fusion and matching of
information in multi-robot systems"

Short abstract

This work introduces a framework for fusing and
matching information in multi-robot systems.  We use
multi-dimension geometric information spaces (loosely
inspired by Gärdenfors's conceptual spaces) to represent the
various types of information used in robotic systems,
including combinations of symbolic/numeric, perceptual/a
priori, and interoceptive/exteroceptive information.
Grounding functions are used to map the various types of
information into corresponding dimensions of the spaces;
these functions can be given or they can be determined
online.  Information from various sources can be represented
in a separate shared space, containing common dimensions and
units.  The shared space can be instantiated in a
centralised or distributed way, and the mappings between the
individual and shared spaces can range in complexity.  From
this shared space, various items of information can be
fused, in order to improve the distributed knowledge in the
system.  As well, information can be matched, in order to
address the anchoring and data-association problems.  We
also describe an implementation of the framework based on
fuzzy logic.


Travel reports

Licentiate seminars


Courses Spring 2016


Last modified on February 2007 by Anne Moe