So far, we have created two Albots, Albot0 and Albot1.
Albot0, the first in the series, was designed to investigate how a creature
could utilize an inexact map to find its way home. A robot with sonar
sensors is used. Sonar readings are noisy and, without error correction, the
map created is highly distorted and inaccurate. This presents us with a
challenge: with such a map, how could Albot0 find its way home? This
challenge is often faced in biological agents with different sensory
capacities. We have created an algorithm for Albot0 to find its way home,
using distance and direction information afforded in its inexact spatial
map. Both pieces of information are typically used in biological agents. One
important lesson learned from implementing Albot0 is that the spatiality in
a cognitive map affords us rich and useful information. This argues against
recent suggestions that the notion of a cognitive map is not a useful one.
For details of this work, see: W.K. Yeap (2011) How Albot0 finds its way
home: A novel approach to cognitive mapping using robots. To appear in
Topics in Cognitive Science
Albot1 was designed to investigate how successive views are integrated to
form a perceptual map. A laser-ranging robot is used. Note that the standard
solution adopted by both robotics and cognitive science researchers is to
transform the co-ordinates of objects in the current view to the next. This
approach is attractive because there exists a simple mathematical process to
do so and new probabilistic algorithms have been developed, extending the
basic approach to cope with errors due to sensors. However, humans are found
not to integrate successive views at the saccade levels and this phenomenon
is known as change blindness. Their cognitive maps are fragmented and
imprecise but if their perceptual map is precise and detailed, why so?
Albot1 was created to investigate how a less detailed perceptual map could
be created without having to integrate successive views and yet still
precise enough to help us orient to unseen locations. Fig. 1 shows a
fragmented perceptual map created by Albot1. For details of this work, see:
W.K. Yeap (2011) A computational theory of humans perceptual mapping.
Accepted for Cognitive Science conference. W.K. Yeap, Md. Z. Hossain, and T.
On the implementation of a theory of perceptual mapping. Submitted to AAAI-2011.
We are currently extending our work to show how a perceptual map could be
computed using vision and how landmarks could be identified and used. We are
also working on using Albots to test infants discover concepts in their
world i.e. a theory of original intentionality.
We are keen to use our new algorithms for practical uses of robots. One such
application is to develop robots that can work autonomously in farms such as
vineyards, spring onions farms and asparagus farms (see Figure 2).
Could robots have original intent?