Due to questionable controls, automatons go on ice-cream attack

Autonomous Robots

March 3, 2006

What results when a seasoned academic with a good standing in the field retires - but before departing collects together the subjects that interest him and publishes them in a book? On the positive side, the guy should know what he is talking about; but, on the other hand, he may be a little off the pace. This book attempts to cover a lot of ground. It is subtitled "From Biological Inspiration to Implementation and Control", and this is the general tenor of an effort to achieve a comprehensive coverage of autonomous robots.

George Bekey expertly brings together a diverse range of subject areas including low-level robot control, robot locomotion and humanoid robots through to the control of multiple robots. Each topic is linked nicely to form a cohesive whole, giving the impression of a well-established field rather than the shirty youth that actually exists.

Some aspects of the book are superb. Bekey has successfully walked a tightrope in providing all-encompassing definitions of "robot" and "autonomy", both of which are traditionally difficult concepts to pin down.

He also gives a snapshot of some commercially available robots, such as the dog Aibo, the vacuum cleaner Roomba and Honda's promotional walking robot, Asimo.

But rather than flowing smoothly, the book shudders forwards in fits and starts. One minute we are in a magazine with a light-hearted coverage of science fiction, yet just a few pages later we are in the bowels of robot algorithms, pitting our wits against strangely disparate mathematical control systems theory, 11 continuous pages of which simply sit there in the book without any connection with the rest of it. Perhaps it is cynical to suggest that Bekey had an interest in control theory years ago. The book's target audience is difficult to discern: is it someone with nothing better to do at the hairdressers or a PhD mathematician?

Bekey flounders on many of the more recent aspects of the subject and is out of his depth in discussing collective robots and artificial intelligence in general, and particularly in the possibilities for intelligent robots of the future. A prime example is the chapter "The future of autonomous robots" - one would be better advised to read Ray Kurzweil's original thoughts rather than Bekey's misinterpretations.

We are told on the book cover that " Autonomous Robots is a comprehensive overview of the subject". It is not. Amazingly, educational robots - Cybot, of which there are more in the world than any other, and Lego Mindstorms - are apparently non-existent; domestic robots hardly appear; and the only research worth its salt seems to be that carried out at the Massachusetts Institute of Technology. At times, the book is astonishingly Americocentric.

But my biggest criticism is of the Elton John-style rose-tinted glasses with which Bekey views the topic. Having successfully defined autonomous robots to include many military machines - for example autonomous fighter planes such as the X-36 and even missiles - Bekey tells us that "some form of high-level control is required to ensure that the robots do not harm any humans". He then suggests that this implies that all such robots will be programmed to comply with Isaac Asimov's laws! Bekey needs to get real. In the first place, it is difficult to conceive how a robot lawnmower can be constructed such that it will cut grass but will not under any circumstances harm a human, while programming a missile not to harm humans seems to defeat the object altogether. Perhaps missiles of the future will simply squirt ice cream at the enemy.

Bekey rightly includes a section on unmanned autonomous vehicles, which he calls "flying robots". The coverage trundles along nicely until we are told: "There is also a class of full-size robotic aircraft. Since such vehicles are intended purely for combat operations, they are not discussed here." A little later he says: "We see the possibility of future military engagements with large numbers of unpiloted vehicles bringing death and destruction to remote locations." Although I am not sure who "we" are, it appears that even Bekey admits that missiles will appear in the future, despite the fact that he seems to have not heard about them just yet.

The US Army's Talon robot, which operates on tracks, is introduced, and we are told how it has been "used for several years to handle mines and clear dangerous ammunition". Almost as an afterthought, we are later informed that "it can also be equipped with a variety of weapons", presumably of the ice-cream-deploying type.

It is initially warming to see a whole chapter on robot learning, including reinforcement learning, Q-learning, neural networks and evolutionary algorithms. Unfortunately, it is all a little dated and falls disastrously between the two stools of offering a comprehensive review of the field and presenting in-depth methodologies of how to achieve different types of learning in robots. It comes over too much as someone writing down the different methods he can recall and not doing the research to provide either type of chapter - what a shame.

As the book is meant to be about autonomous robots, chapter ten does not seem to fit in. It contains mainly a traditional approach to robot manipulators, including kinematics and dynamics, that one would expect from a general introductory book on robotics. Sadly, chapter 11 - which examines the "Control of grasping in human and robot hands", apart from a couple of references, classified as "other prosthetic hands" - suffers much the same fate. In both cases, it appears that Bekey had quite a bit of material lying around from the past and wanted to publish it somewhere.

To be fair, Bekey has pulled together a lot of useful material, and the book contains many appropriate photographs and illustrations. Mostly, the robots are presented in an easy and chatty style for a general readership.

It is a shame that Bekey did not complete his task and carry out a thorough and more all-encompassing review along these lines. Had he done so, and left out the general control theory and the futuristic, philosophical sections that are clearly not his forte, this could easily have been quite a good book. There are still those damn rose-tinted glasses, though.

Kevin Warwick is professor of cybernetics, Reading University.

Autonomous Robots: From Biological Inspiration to Implementation and Control

Author - George A. Bekey
Publisher - MIT Press
Pages - 577
Price - £35.95
ISBN - 0 262 02578 7

Register to continue

Why register?

  • Registration is free and only takes a moment
  • Once registered, you can read 3 articles a month
  • Sign up for our newsletter
Register
Please Login or Register to read this article.

Sponsored