Summary: 19 Mar 2002 3:48 p.m.
Data Fusion Algorithms for Collaborative Robotic Exploration
Jeremy Thorpe and Robert McEliece.
In this paper we will study the problem of efficient data fusion in an ad hoc network of
mobile sensors ("robots") using belief propagation on a graphical model similar to those
used in turbo-style decoding. We also devise a new metric for evaluating the performance
of general inference algorithms, including BP, which return "soft" estimates.
1. Introduction. The Robot Inference Problem.
In the future, NASA spacecraft sent to extraterrestrial planets to collect scientific data
may deploy a large number of small, inexpensive robots. Each of the robots may be
equipped with a number of sensors with which to measure its local microenvironment.
Individually, one robot's measurements would convey little information about the global
situation. Collectively, however, the robots could overcome this by communicating with
their neighbors and fusing their data. The robots should thus be able to collectively infer a
great deal about the global environment. The trick is to do it with as little communication
and/or computation as possible.
In this paper, we will begin to explore the problem of designing efficient data fusion
strategies in an ad hoc network of robotic sensors, by first defining a simplified model for
the problem (Section 2), and then applying the celebrated belief propagation algorithm