Biorobotics: Methods & Applications

Citation

Andrew, A.M. (2002), "Biorobotics: Methods & Applications", Kybernetes, Vol. 31 No. 6. https://doi.org/10.1108/k.2002.06731fae.007

Biorobotics: Methods & Applications

Barbara Webb and Thomas R. Consi (Eds)AAAI PressMenlo Park, CA and MIT Press Cambridge MA2001 (distributed by MIT Press)xiv + 208 pp.ISBN 0-262-73141-Xpaperback, £23.95

This is an edited version of papers that were presented at an American Association for Artificial Intelligence (AAAI) Symposium in Orlando, Florida in October 1998, with the topic: "Robots and Biology: Developing Connections". The book has eight papers, as well as Introduction and Epilogue by the editors, and a common bibliography and subject index.

Any development in robotics invites, almost by definition, comparison with some biological system, but in this new field considerations of animal and machine are combined particularly intimately. Its development is greatly encouraged by the availability of small and low-power information processors and sensors and actuators. As the editors say in their Introduction, biorobots have dual uses since they can be tools for biologists studying animal behaviour and can also be testbeds for the study of biological algorithms with potential applications to engineering. The contributions in this book are concerned mainly with the role as tools for biologists.

It is possible to question the value of building physical robots rather than using computer simulations. The justification comes from consideration of the complexity of the sensory world with which living creatures interact. A computer simulation must be an approximation whose adequacy can never be assured. One of the papers, for example, describes a biorobot that models the use of olfaction by lobsters and other underwater creatures to locate the source of an odorous substance conveyed to them through turbulent water. Diffusion through this medium, especially if disturbed by movements of the lobster itself, is very difficult to simulate, and a model operating in real water is an obvious solution.

Another reason that the building of physical robots is instructive is that it allows the exploration of parsimonious ways of achieving the required performance, and particularly of ways that have some correspondence to the biological prototype. Parsimony is particularly relevant to studies of insect behaviour since insects achieve complex behaviour with remarkably few neurons.

The interactions of biorobots with complex environments may also give insight into how higher cognitive functions of animals have evolved, and into adaptation as part of higher brain function.

Of the eight papers in the book, four deal primarily with sensory systems, two with motor systems and two with cognitive systems. The distinction between the first two groups is somewhat arbitrary since all the schemes depend on interaction of sensory and motor functions, but with one or other judged to be the main focus of interest.

Of the four papers in the first group, one is about a robot simulating a female cricket guided towards a particular male by the sound he makes, irrespective of other sources of sound including other males. The next is about the olfactory capabilities of lobsters, already mentioned, and another is a simulation of visual homing by insects, especially desert ants which cannot use chemical markers in their arid surroundings. The fourth paper is about visual stabilisation of a robot using propellers to simulate insect wings, and achieving remarkable stability by very simple means.

The first paper in the motor systems group refers to a robot simulating the locomotion and postural mechanisms of the cockroach, and the second is about control of complex motor systems in general, again with a reference to a six- legged cockroach-like model. The suggestion arises here that the origins of the cognitive functions of higher animals can be found in the integrative processes needed for motor control.

The first paper of the cognitive systems group explores evolutionary aspects from a different viewpoint and considers how principles of invariance of visual patterns could arise from visual experience of an active animal or robot. The other paper refers to the study of social development using the humanoid robot ''Cog'' developed by R.A. Brooks at MIT. Cog can locate the face, and the eyes, of a human confronting it and can determine whether the human is looking at him, and can return the gaze. He (or it) can also perform arm movements that can be imperative or declarative and can imitate actions such as nodding.

The Epilogue gives a thoughtful discussion of the place and usefulness of biorobotics. That the approach can give useful insights is demonstrated by the first two papers in the book, where the first shows that aspects of cricket phonotaxis can be explained much more simply than was originally supposed, whereas the second shows that location of the source of an odorous plume in turbulent water is more difficult than was originally supposed, and requires the use of multiple strategies by the lobster. Despite the diversity of the contributions, the participants of the symposium felt they shared common ground.

It is of course important to be realistic about the nature of biorobotic experiments. A biorobot is designed to test one particular aspect of a living system, and simulates it only in certain respects. There are, in fact, no relevant living systems that have been analysed in sufficient detail, including their neural underpinning, to permit complete modelling. Nevertheless a biorobotic approach can be said to have ''arrived'' when it generates biorobots that can be subjected to precisely the experimental procedures that a biologist would apply to a living system.

Progress in the area is hampered by a shortage of engineers knowing enough biology and of biologists knowing enough about engineering. It is also hampered by lack of a recognised forum for publications, which at present are widely scattered. On the other hand the setting up of a specialist journal or conference series is not felt to be appropriate since it could tend to isolate the field from the main streams of biology and engineering from which it draws its strength and justification. A website and mailing list resulting from the AAAI symposium can be found at: http://www.ai.mit.edu/~scaz/biorobotics/ and will receive attention in an Internet Commentary.

The book contains a great deal that is intriguing and thought-provoking and in the true interdisciplinary spirit of cybernetics.

Alex M. Andrew