SAB '02 WORKSHOP
ROBOTICS AS THEORETICAL BIOLOGY
Robot III (Roger Quinn, CWRU)
AUGUST 10TH, 2002
workshop email address
Part of SAB '02:
The 7th Meeting of the International Society
for Simulation of Adaptive Behaviour
August 4-11th, 2002.
with additional sponsorship from The Wellcome Trust
*** THE WORKSHOP TOOK PLACE AS PLANNED WITH OVER 40 PARTICIPANTS INVOLVED ***
*** THANKS TO ALL THOSE WHO TOOK PART ***
This workshop seeks to bring together researchers in robotics and biology who are
interested in developing robot models of the biological systems underlying animal behaviour.
The scope of the workshop will include:
- Evaluating current progress in applying robot models to biological questions
- Identifying areas of biology where robots could make a future contribution
- Exploring methodological issues
- Considering how better ties can be forged between the robotics and biological research communities
- Providing a forum for exhibiting current research and work in progress
FORMAT OF THE WORKSHOP
This will be a one-day workshop consisting of a mixture of invited talks and discussion sessions (see PROGRAM below).
The workshop will include three sessions of invited talks: one on invertebrate behaviour,
one on vertebrate behaviour, and one on collective behaviour. For each session there will be two invited speakers,
one a distinguished researcher from a (primarily)
biological background, the other from a robotics background. Each invited speaker will be asked to
address the question "How can robotics contribute to theoretical biology?"
from the perspective of their own field and drawing on the experience of their own research.
Each session will also conclude with a discussion involving the two speakers
and a panel of other prominent researchers in the field. The workshop will end
with a discussion about building stronger ties between the biological and
robotics research communities. Representatives from funding bodies
and from journal editorial boards will be asked to join the panel for
All attendees at the workshop are invited to bring posters
describing their own research for people to browse during scheduled coffee and lunch breaks
(see CONTRIBUTING TO THE WORKSHOP below).
The workshop is co-organised by:
Tony Prescott, Adaptive Behaviour Research Group, University of Sheffield, UK, and
Barbara Webb, Department of Psychology, University of Stirling, UK.
Chair: Barbara Webb
Talk 1: Roy Ritzmann (Professor of Biology, Case Western Reserve University, US)
Solutions to Moving over Complex Terrain Found in Cockroach Locomotion (Abstract)
Talk 2: Roger Quinn (Professor of Mechanical Engineering, Case Western Reserve University, US)
Parallel Strategies For Implementing and Testing Biological Principles In Mobile Robots (Abstract)
Discussion Panel: Roy Ritzmann, Roger Quinn,
Chair: Tony Prescott
Talk 3: Peter Redgrave
(Professor of Neuroscience, University of Sheffield, UK)
Neuroscience, Robotics, and the Shared Problem of 'What to do Next' (Abstract)
Talk 4: Angelo Arleo (Research Fellow, College de France, Paris, France)
Processing multimodal sensory information for
spatial learning and navigation: Computational modeling, robotics, and rat experiments (Abstract)
Discussion (Panel to be confirmed)
Chair: To be confirmed.
Talk 5: Nigel Franks (Professor of Biological Sciences, University of Bristol).
Collective Intelligence in Ants (Abstract)
Talk 6: Chris Melhuish
(Director of the Intelligent Autonomous Systems Laboratory, University of the West of England, UK).
Autonomy in Collective Minimalist Robot Systems (Abstract)
Discussion (Panel to be confirmed)
Chair: To be confirmed
Discussion (Panel to be confirmed but to include representatives from research funding bodies, research societies, and journal editorial boards)
Panel so far:
Sue Healy one of the editors of the journal
Jay Rosenberg one of the editors of the new
Journal of Integrative Neuroscience.
Vicky Jones and John Hand of the UK Engineering and Physical Sciences Research Council (EPSRC).
Laura Playle of the UK Biotechnology and Biological Sciences Research Council (BBSRC).
Jean-Arcady Meyer, co-founder of ISAB
and of the journal Adaptive Behavior.
CONTRIBUTING TO THE WORKSHOP
All attendees are invited (but not required) to bring posters, and robot demos, of relevant research to the workshop.
Posters may have been previously exhibited at other recent workshops/conferences (including SAB '02),
or may describe new research that has not been exhibited elsewhere.
If you wish to show your work
you are requested to submit an abstract of 200-400 words to the workshop organisers before the 15TH OF JUNE 2002.
Submissions will not be reviewed however they will be vetted for relevance,
and work that falls outside the scope of the workshop may be refused. Potential contributors
should note that the focus of the workshop is on the possible contribution of robotics
to biology, rather than the converse (the contribution of biology to robotics).
Posters describing biologically-inspired robots, computer simulation work, or straight biological
research are acceptable where they can be shown to have relevance to the understanding or development
of future robot models in biology. Please make the biological relevance clear in your abstract.
Attendees who wish to exhibit robots should also submit an abstract and should contact
the organisers with full details of the demonstration and any technical requirements.
There may be a limit on the number of available poster spaces, and there may also be space restrictions on
robot demonstrations. If these limits become an issue then contributions will be prioritized according
to the order in which they were received.
Please note, we are NOT inviting submissions for oral presentations, however,
the program is organised for lots of discussion which will certainly be open to
input from all attendees.
For enquiries and for electronic submission of abstracts
email the workshop organisers.
There will be a proceedings booklet for attendees that will include the abstracts of invited talks and
accepted posters/demos, contact details for all attendees, and other pertinent information. Attendees
with accepted abstracts will be asked to contribute 1 page of 'camera-ready' material to the proceedings
(formatting instructions to be published here at a later date).
Please book through the SAB 'O2 booking page. Note: it IS possible to register
for the workshop without registering for the full conference.
Workshop email address
SAB '02 homepage
Biorobotics resource page
Biomorphic robot links
Can robots make good models of biological
behaviour?, BBS 24(6), 2001.
Biorobots: Methods and Applications
by Barbara Webb and Thomas Consi, MIT Press, 2001.
ABSTRACTS OF INVITED TALKS
Although wheeled vehicles move efficiently over level ground or minor gradations,
they cannot match the agility of legged animals to traverse the complex terrains
found in nature. We believe that it was this challenge that provided the selection
pressures that led to the complex leg structures and related motor systems found
in modern animals. These are also the very abilities that make legged locomotion
attractive to robotic designs. With this in mind, we have chosen to investigate
what we call "transitional" behaviors that take animals, and in particular insects,
over, around and under barriers of varying physical properties. Our observations
suggest that these behaviors require an interplay between local control and higher
centers of the central nervous system that allow the animal to detect and evaluate
an obstacle well before reaching it. The result is a smooth transition from forward
locomotion to climbing, turning or tunneling. For example, in climbing a substantial
block, the cockroach first rears up and places its legs on top of the obstacle.
It typically does this without contacting the front of the barrier with any leg,
suggesting that it has measured the block with sensors such as antennae. The actual
movement of the body upward requires an increase in motor activity that is probably
adjusted by local reflexes. After lesioning various brain regions, the animal can
still move forward, but cannot alter its trajectory in anticipation of obstacles.
Rather they can only react after hitting the barrier. We are currently examining
the linkage between higher centers and local control circuits that generate the
smooth transitions found in intact animals.
(Back to Session I Programme)
Our goal is to use intelligent biological inspiration to develop robots that capture the capacity
of animals to traverse complex terrain. We follow two distinct but complementary strategies to meet
this goal. In one, we produced a series of robots that have mobility increasingly more similar
to that of a cockroach. The leg designs of these robots assure that they can generate movements
used by the cockroach to walk and climb over a range of objects. However, in order to take
advantage of these complex designs, we must first solve difficult problems in actuation, proprioception
and control and in solving these problems we learn more about the animal. The second parallel
strategy seeks to capture the principles of biological movement, but in an abstract form that does
not require complex platforms. Following the second strategy, we designed and built two new robots that
use one drive motor to generate a nominal tripod gait. Gait changes similar to those used by the animal
are accomplished through passive mechanisms and rearing movements in anticipation to climbing are
accomplished by way of a body flexion joint. The parallel development of these robotic lines provides
the best of both worlds. The complex designs will ultimately be more versatile and agile than the
simplified line, but will take more effort to perfect. The simplified line provides short-term solutions
that can be deployed immediately and confirm, in principle, the value of biological properties for
complex locomotion. Furthermore, these robots can be used immediately to implement and test sensor
and navigation strategies extracted from animals. Please see the
CWRU Biorobotics homepage
for more details.
(Back to Session I Programme)
Any mechanism, biological or mechanical, which has more than one sensory or cognitive
representation capable of directing action faces the problem of 'what to do next'.
To avoid chaotic movement, potential conflicts between independent functional
systems must be resolved and control over the 'final common motor path'
distributed in an orderly manner. This issue, often termed action selection,
continues to be a major concern of ethology, psychology, neuroscience, and robotics.
Our research group has recently proposed that part of the vertebrate central
nervous system, the basal ganglia, appears to be ideally configured to select
between multiple cognitive and sensori-motor systems which have the capacity
to promote exclusive behaviours. To test this notion we have implemented a
high level computational model of intrinsic basal ganglia circuitry and its
interactions with other regions of the brain. The model was then exposed to
the rigors of 'real world' action selection by embedding it within the control
architecture of a small mobile robot. In this talk, this recent basal ganglia
research will be used as a context for considering the potential contribution
of robotics to the neurosciences, and the prospects for future collaborations
between researchers in these two fields.
(Back to Session II Programme)
Hippocampal place (HP) cells and head direction (HD) cells are a likely neural basis for spatial
orientation in rats. A HP cell fires maximally only when the rat is at a specific location.
HD cells discharge as a function of the rat's allocentric heading, regardless of the rat's
location and behavior, or of the orientation of the head relative to the body.
We study the mechanisms underlying the location- and direction-selectivity of HP and HD cells,
respectively. Here, we discuss a computational model in which place and directional coding are
provided by two neural networks that interact with each other to form a unitary spatial
learning system. The model stresses the importance of integrating multimodal sensory signals
to establish robust place and direction representations. Thus, the dynamics of both neural
networks relies on allothetic (visual) and idiothetic (egomotion) information.
The place-coding network models the rat hippocampal formation and involves, in particular,
the superficial layers of the entorhinal cortex, the subiculum, and the CA3-CA1 regions.
Unsupervised learning is applied to extract spatio-temporal properties of the environment
from visual input. Path integration is used to build an idiothetic map based on self-motion cues.
Then, long-term potentiation (LTP) and long-term depression (LTD) are used to combine
the allothetic and idiothetic representations.
The direction-coding network models HD cells observed in the anterodorsal thalamic nucleus (ADN),
lateral mammillary nuclei (LMN), and postsubiculum (poSC). LMN and poSC are modeled by attractor
networks. Idiothetic signals control the dynamics of ADN and LMN, whereas allothetic
cues mainly influence poSC cells and modify the system's dynamics by occasionally
reorienting the directional representation.
We validate the theoretical model on a mobile robotic platform and employ experimental
protocols similar to those used for rat experiments.
The objective of this approach is to develop
a theory compatible with experimental data and to provide predictions (concerning
neurophysiological, anatomical, and behavioral issues) that could lead to innovative
experiments with animals.
(Back to Session II Programme)
I will consider why rather few insights have, as yet, flowed back from the study of societies of distributed robots to the ant biology that,
in part, inspired their creation. One reason that insights and inspiration have been largely unidirectional is that certain misconceptions
seem to be commonplace. There are 3 major misconceptions. First, that ants are simple and stupid. Second that self-organization is an
alternative explanation to natural selection for complexity in biology. Third, that in these systems there is "order for free". These
misconceptions are, more or less strongly, linked to one another.
First, I will show that ants are neither simple nor stupid, either individually or collectively. Second, I will present evidence that
natural selection has had a key role in the self-organization of certain ant behaviours and hence I will show that the term
"order for free" is misleading.
I will illustrate these points by briefly considering two very different types of ants - army ants and rock ants.
I will then point out that a major challenge is to understand how ants, with dazzling sophistication at the individual
and collective levels, prioritise and organize their multi-faceted lives. This includes, brood sorting, worker sorting,
building behaviour, emigrations, orientation, nest site assessment and quorum sensing in house hunting.
(Back to Session III Programme)
From an engineering perspective, the design and construction of mobile autonomous systems
often draws inspiration from the natural world - the existence proof of truly intelligent
autonomous systems. Autonomous robots will be exposed to many of the problems faced by
biological systems. It is therefore interesting for roboticists to understand the
mechanisms exploited by natural systems.
In particular, the work on minimalist collective robot systems at the
Intelligent Autonomous Systems Laboratory (IAS)
was inspired by the sorting behaviour of Lepothorax ants. From this engineering
perspective the question posed was how 'simple could a robot controller be?'
and still provide robustness, redundancy and low unit cost. In this context
I will discuss how simple rule sets have been developed to achieve 2 and 3
object sorting and how the use of simple rules with temporal augmentation
(ie the use of leaky integrators) have been employed to generate annular sorting.
Naturally, one might ask if these rules are seen in Nature. Currently, the answer seems
to be 'unlikely' - but there is much to learn about how Leptothorax tackles the problem;
the full picture is far from understood. A joint project between roboticists and
biologists has recently started in the IAS lab to look at this problem. Some initial findings
will be discussed as well as observations on the symmetry of the flow of insights and
questions between the two disciplines.
(Back to Session III Programme)
[an error occurred while processing this directive]