Berthouze, L., Kaplan, F., Kozima, H., Yano, H., Konczak, J., Metta, G., Nadel, J., Sandini, G., Stojanov, G. and Balkenius, C. (Eds.)
Proceedings of the Fifth International Workshop on Epigenetic Robotics: Modeling Cognitive Development in Robotic Systems
Lund University Cognitive Studies, 123. ISBN 91-974741-4-2
Ongoing Emergence:
A Core Concept in Epigenetic Robotics
Christopher G. Prince, Nathan A. Helder, & George J. Hollichψ
Department of Computer Science
University of Minnesota Duluth
Duluth, MN 55812 USA
[email protected], [email protected]
ψPurdue University
Department of Psychological Sciences
West Lafayette, IN 47907 USA
[email protected]
Abstract
We propose ongoing emergence as a core concept in
epigenetic robotics. Ongoing emergence refers to the
continuous development and integration of new skills
and is exhibited when six criteria are satisfied: (1)
continuous skill acquisition, (2) incorporation of new
skills with existing skills, (3) autonomous development
of values and goals, (4) bootstrapping of initial skills, (5)
stability of skills, and (6) reproducibility. In this paper
we: (a) provide a conceptual synthesis of ongoing
emergence based on previous theorizing, (b) review
current research in epigenetic robotics in light of ongoing
emergence, (c) provide prototypical examples of ongoing
emergence from infant development, and (d) outline
computational issues relevant to creating robots
exhibiting ongoing emergence.
1. Introduction
Epigenetic robotics is a new field that focuses on
modeling cognitive development and creating robots that
show autonomous mental development (Lungarella,
Metta, Pfeifer, & Sandini, 2003; Weng, McClelland,
Pentland, Sporns, Stockman, Sur, & Thelen, 2001). For
example, robots have been implemented that generate
visual discrimination behavior using large-scale neural
networks (Seth, McKinstry, Edelman, & Krichmar,
2004), that model early infant-caregiver interaction using
behavioral rules (Breazeal & Scassellati, 2000), and that
explore the knowledge needed by infants to succeed in
perceptual object permanence experiments (Chen &
Weng, 2004; Lovett & Scassellati, 2004; see also:
Schlesinger & Casey, 2003). Given these and other
diverse contributions to this new field (for a review, see
Lungarella et al., 2003) it seems an opportune time to
synthesize a few core concepts from this corpus of
research.
In this paper, we distill one such core concept,
ongoing emergence, which refers to the continuous
development and integration of new skills. An agent
exhibiting ongoing emergence, in a motivationally
autonomous manner, will continue develop and refine its
skills across development. This vision of open-ended
development is evident in recent work. For example, in
efforts to “allow a mobile robot to incrementally progress
through levels of increasingly sophisticated behavior” (p.
1, ms., Blank, Kumar, Meeden, & Marshall, 2005), in
efforts to build robots that exhibit “new behavior, which
in turn, becomes a precursor for successive stages of
development” (p. 27, Grupen, 2003), and in efforts to
achieve robots exhibiting a “successive emergence of
behaviors in a developmental progression of increasing
processing power and complexity” (p. 1, ms., Dominey &
Boucher, 2005). Unfortunately, while humans clearly
show such long-term progressions, epigenetic robots as
yet do not—they are typically designed to achieve
particular behaviors or to learn specific tasks.
To escape this impasse, we propose a theoretical
framework for achieving ongoing emergence. To this
end, in Section 2 we review previous theoretical
conceptions regarding ongoing emergence and synthesize
the current state of the art in terms of six criteria. Section
3 considers how current examples of robotic systems fare
with respect to these criteria for ongoing emergence. In
Section 4, we look to infant developmental research for
examples of ongoing emergence. Section 5 outlines some
computational issues for designing robots that exhibit
ongoing emergence. We close with a discussion.
2. Conceptual Synthesis
2.1. Background
Blank et al. (2005) discuss the possibility that a robot can
use a developmental algorithm to learn, via a process of
self-exploration, its repertoire of behaviors and mental
capabilities, instead of being preprogrammed with “the
capabilities of a human body and human concepts” (p. 2,
ms). Robots are proposed to discover even the most
primitive behaviors through a process of exploration.
A possible benefit of providing such a developmental
algorithm is avoiding specification of task-goals for the
robot. Instead, “it is the goal of developmental robotics to
explore the range of tasks that can be learned (or grown)
by a robot, given a specific developmental algorithm and
a control architecture” (p. 2, ms). These authors consider
three mechanisms to be essential to developmental
algorithms: abstraction, anticipation, and self-motivation.
Abstractions are seen as necessary to focus the robot’s
attention on relevant environmental features, given the
“constant stream of perceptual information” (p. 2, ms.).
Anticipations enable the robot to predict environmental
change to “go beyond simple reflexive behavior to
purposeful behavior” (p. 3, ms.). And self-motivation
“push[es] the system toward further abstractions and
63