October 23-35, Orlando Florida
This symposium proposes a focused topic for discussion by a diverse group of participants. The aim is to present and promote work in the use of robot models to test biological hypotheses. But we wish to attract researchers currently involved in any part of that equation: building robots; doing biological modelling; taking biological inspiration for engineering; or taking robotic inspiration for biological hypotheses. Major aims would be to systematically explore when and how robots can be used in biology, when biological ideas can be applied to robotics, and what both tell us about intelligent control of behaviour.
Additional aims are:
The proposed format of the symposium includes:
i) Set the scene: short presentations and discussions of existing work in this area (see below) to establish what is currently possible and raise basic issues, especially of methodology
i) Find the connection: participants would be asked to provide in advance a short summary of their current work, focussing on what they are now looking for, e.g. the means for testing by modelling a certain biological hypothesis; a biological analog of a required robot capability; etc. Time would then be provided for people to meet in small groups to cover common ground on specific issues, and potentially establish new collaborations.
iii) Spot the difference: a panel discussion of advantages and limitations in drawing close connections between robotics and biology; and what advances are required in each field to support futher interaction.
iv) Draw the conclusions: a panel discussion of what new insights on the problem of building artifical intelligence is being provided by these low-level approaches.
v) Point the way: prepare summary documents identifying new areas of research in biology and robotics where a hybrid approach may be used, and potential developments that would support this interaction
There is clearly common ground between robotics and biology in trying to understand how sensorimotor mechanisms can support adaptive behaviour. But what does this suggest in terms of fruitful modes of interaction between these fields? Despite substantial growth in the last decade in the sub-field of 'behaviour-based' AI (the study of non-intellectual, sub-linguistic intelligence using mechanisms 'inspired by' biology) most links have remained at a rather abstract level.
However there is a rapidly expanding body of work that draws very explicit connections. Examples include: the fly-vision model for robot obstacle avoidance (Franceschini et. al); simulated and physical models of six-legged insect walking (Cruse et.al.; Beer et. al.); robot modelling of cricket phonotaxis (Webb, et. al.), lobster chemotaxis (Consi et. al.) and polarised light navigation (Wehner et. al.). There are a number of other systems in development. At the same time there is also growth in closely related work in computational neuroethology (i.e. software models eg. Cliff, Arbib), neuromorphic engineering (biologically based hardware designs, Mead) and neural-network controlled robotics.
The possibility of building electronic and computational systems that closely model biology is becoming increasingly viable:
Understanding of how biological systems function is advancing rapidly, particularly at the level of neural underpinning of simple behaviours. There are a number of biological systems for which a great deal of information about their sensorimotor mechanisms is available, much of which is yet to be exploited.
Neuroethologists are becoming increasingly interested in the potential power of physical modelling for testing hypotheses about the function of biological systems. Physical models can have advantages over software simulation because they can operate in natural environments and be challenged with real, complex stimuli.
The availability of inexpensive yet powerful microprocessors, solid state sensors and miniature electromechanical components has made possible the construction robots that can, to a limited degree mimic certain features of animals. These biomimics can be used as surrogate animals to study the sensory or motor systems that have been copied.
Engineers are becoming increasingly interested in the potential power of biologically inspired designs as novel means for building sensors and controllers
Researchers in AI are becoming increasingly convinced that the foundations of intelligence lies in the control of behaviour, and that this can be established by having properly crafted sensors and actuators that vastly simplify the intervening processing required.
Nevertheless the diversity of the field means that currently there is no single forum for bringing together researchers from this range of backgrounds. Discovering examples of research in this area, or important connections between approaches, is too often a matter of chance. Sessions and workshops have tended to lack sufficient focus to enable real, co-operative progress to be made.
This symposium aims to advance work in this promising field by addressing the following issues:
Potential participants should submit a short statement (no more than two pages) that
i) Gives your name, affiliation, mailing address, email address, phone and fax numbers
ii) Summarises your previous and current work relevant to this field, with pointers to further information in published papers and/or on websites
iii) Provides a list of what issues you would like to see discussed, what questions you want answered, and what connections you would like to make.
Submissions should, by preference, be sent in plain text email to Barbara.Webb@nottingham.ac.uk with "Robotics and Biology:submission" in the subject line. Hard copy submissions should be 4 copies sent to: Barbara Webb, Department of Psychology, University of Nottingham, Nottingham, NG7 2RD, U.K.
Deadline: April 15th 1998
Barbara Webb, AI Group, Dept. of Psychology, University of Nottingham Barbara.Webb@nottingham.ac.uk
Thomas Consi, MIT Dept. of Ocean Engineering, consi@mit.ed
Holk Cruse, Faculty for Biology, University of Bielefeld, holk@bio128.uni-bielefeld.de
Randall Beer, Dept. of Computer Engineering and Science & Dept. of Biology, Case Western Reserve University, beer@alpha.ces.cwru.edu