17-03-2014, 06:49 PM
Abstract
We present a formal method, based on the Logic of
Possibility, to fuse uncertain senso y information and
to produce an estimate of the position of a mobile robot.
The robot navigates in an ofice environment, using a
topological map, with the assistance of a “slave” robot
acting as a portable local landmark. Each relevant place
in the map is characterized by a set of logical formulae
axiomatizing both “crisp” knowledge and uncertain
information from the sensors. At each time instant
during navigation, the necessity degree of each place is
calculated using a purely syntactical method based on,
sequent calculus.
1 Introduction
This paper elaborates on the ideas presented in previous
work [1, 21. There we proposed a feature-based
localization technique for navigation in a partially observable
office environment. The approach uses a topological
map augmented with additional information, including
approximate metrics. As usual in this class of
approaches, at every time step, an estimate is produced
of the location of the robot in the map, in the form of
a certainty degree distribution, over a set of relevant
places. The actual localization then consists in interpreting
such distribution by assuming a preferred state
according to some criterion. Our experimental testbed
consists of a team of two cooperating robots navigating
in an office area, with one robot using the other as a
local visual landmark.
Our goal is to investigate on the adequacy of Possibilistic
Logic [7], in an enhanced version (LPL) proposed
by our own group [4], for the treatment of uncertainty
in localization, as an alternative to other current
approaches, mainly probabilistic.
The application of our formalism concerns two aspects:
0 the coding and the update of the certainty distri-
0 the integration of heterogeneous sources of infor-
With respect to our previous papers, we have int>ro-
0 we consider a slightly more complex environment,
bution over the map,
mation.
duced some significant additional elements:
with two rooms and a connecting corridor;
0 therefore we cannot rely only on sensory signatures
for localization, but we need to integrate them
with higher level abstract knowledge;
0 and we need to express our axioms in predicative,
rather than propositional form.
We present a formal method, based on the Logic of
Possibility, to fuse uncertain senso y information and
to produce an estimate of the position of a mobile robot.
The robot navigates in an ofice environment, using a
topological map, with the assistance of a “slave” robot
acting as a portable local landmark. Each relevant place
in the map is characterized by a set of logical formulae
axiomatizing both “crisp” knowledge and uncertain
information from the sensors. At each time instant
during navigation, the necessity degree of each place is
calculated using a purely syntactical method based on,
sequent calculus.
1 Introduction
This paper elaborates on the ideas presented in previous
work [1, 21. There we proposed a feature-based
localization technique for navigation in a partially observable
office environment. The approach uses a topological
map augmented with additional information, including
approximate metrics. As usual in this class of
approaches, at every time step, an estimate is produced
of the location of the robot in the map, in the form of
a certainty degree distribution, over a set of relevant
places. The actual localization then consists in interpreting
such distribution by assuming a preferred state
according to some criterion. Our experimental testbed
consists of a team of two cooperating robots navigating
in an office area, with one robot using the other as a
local visual landmark.
Our goal is to investigate on the adequacy of Possibilistic
Logic [7], in an enhanced version (LPL) proposed
by our own group [4], for the treatment of uncertainty
in localization, as an alternative to other current
approaches, mainly probabilistic.
The application of our formalism concerns two aspects:
0 the coding and the update of the certainty distri-
0 the integration of heterogeneous sources of infor-
With respect to our previous papers, we have int>ro-
0 we consider a slightly more complex environment,
bution over the map,
mation.
duced some significant additional elements:
with two rooms and a connecting corridor;
0 therefore we cannot rely only on sensory signatures
for localization, but we need to integrate them
with higher level abstract knowledge;
0 and we need to express our axioms in predicative,
rather than propositional form.