Behavior‐Based Robotics

Citation

Andrew, A.M. (1999), "Behavior‐Based Robotics", Kybernetes, Vol. 28 No. 1, pp. 107-108. https://doi.org/10.1108/k.1999.28.1.107.3

As the notes on the dustcover say, this is the first true survey of this robotics field. It covers a very wide range of topics, all treated very clearly and outlining and contrasting diverse approaches and viewpoints. The reference to behaviour (with or without the letter “u”) carries two implications ‐ first that the concern is primarily with the “brains” of robots rather than with their physical features, and secondly that there is frequent reference to relevant theories and models from biology and psychology. Of course, robotics is by its very nature a topic where the “brains” cannot be considered in isolation from the physical interactions, so the primary concern with perceptual and control processes must be mainly a matter of emphasis.

The foreword by Michael Arbib is enthusiastic, not least about the way the presentation is embellished with apposite quotations from a wide variety of sources, and about the copious provision of illustrations. Arbib makes the slightly laboured but nonetheless apposite comment that although the primary concern is with robot brains, the book gives many opportunities to admire human wit and robot bodies.

The author explains that he realised the need for the book, and decided to write it, when he was unable to find a suitable text for a course he had taught for ten years. He had been forced to recommend to the students rather indigestible technical material and collections of original papers. The book’s intended audience includes upper‐level graduate students and graduate students studying AI and robotics, as well as those interested in learning more about robotics in general.

Developments in robotics are covered over a long time span. Attention is restricted to the electronics era, so there is no review of the various golems and such things as Vaucanson’s rather offputting duck. There is a full description of Grey Walter’s tortoise, and the story is continued right to the latest developments. Michael Arbib comments favourably on the inclusion of motor schema theories, and there are also references to subsumption architecture, and to reinforcement learning in its modern robotics context, among much else.

Although the book has been prepared primarily as a teaching text, and will serve that purpose admirably, it is written with a good deal of flair and imagination. This is perhaps best illustrated by quoting the last three paragraphs of Arbib’s Introduction:

The book’s chapter on social behavior presents what is itself a relatively recent chapter of robotics ‐ sociorobotics. While in its infancy, we can see in the studies of robot teams, inter‐robot communication, and social learning the beginnings not only of a powerful new technology, but also of a new science of experimental psychology.

Finally, we are taken to that meeting place between science fiction, philosophy, and technology that attracted so many of us to wonder about robots in the first place. The final chapter, “Fringe robotics: beyond behavior” (a nod to the 1960s British review “Beyond the fringe”?), debates the issue of robot thought, consciousness, emotion and imagination, returns to Arkin’s long‐standing concern with the possible utility to robots of analogs of hormones and homeostasis, and closes with an all too brief glimpse of nanotechnology.

In this way we are given a tour that impresses with the depth of its analysis of the schemas underlying robot behaviour, while continually illustrating the deep reciprocity between robotics and biology, psychology, sociology, and philosophy, and the important connection between robotics and many other areas of computer science. This is a subject whose fascination can only increase in the decades ahead as many researchers build on the framework so ably presented here.

Ashby’s homeostat receives honourable mention in the context of hormones and homeostasis. Under the heading of nanotechnology there is mention of a Japanese project which has produced a system that is part robot and part cockroach, in that two severed cockroach legs are controlled by a microprocessor. The roboroach has a robot body but biological legs, which are said to remain operational for about an hour.

Arbib’s glowing commendation is fully deserved. There is a comprehensive subject index and name index as well as a valuable bibliography of no less than 32 pages. This is a valuable reference work as well as a thought‐provoking and imaginative review, at the same time undoubtedly serving its initial purpose as a teaching text.