Embedded Robotics: Mobile Robot Design and Applications with Embedded Systems

Kybernetes

ISSN: 0368-492X

Article publication date: 1 June 2004

103

Citation

Andrew, A.M. (2004), "Embedded Robotics: Mobile Robot Design and Applications with Embedded Systems", Kybernetes, Vol. 33 No. 5/6, pp. 1066-1067. https://doi.org/10.1108/03684920410534137

Publisher

:

Emerald Group Publishing Limited

Copyright © 2004, Emerald Group Publishing Limited


This is a detailed account of a very comprehensive and successful study of robotics, conducted mainly in the University of Western Australia, but with acknowledgement of the participation of a number of other centres, in Canada, Germany, New Zealand and USA. The assistance of colleagues in various locations in writing the book is acknowledged. The main development is a remarkably powerful controller, termed “EyeCon”, that has been produced in a suitable form to be installed in quite small robots. When connected to a digital camera it is sufficiently powerful to allow on‐board image processing, and it has the means of interacting with other sensors and actuators with high precision. It can run programs in C and C++, and so complex robot behaviour is readily achievable.

The EyeCon controller itself, as well as associated hardware, and a variety of complete robots termed the EyeBot family, are available commercially from a company called Joker Robotics, with Web site: http://joker‐robotics.com. The EyeBot family includes vehicles of various kinds as well as six‐legged walkers, biped android walkers, and a flying robot. All the software mentioned in the book is freely available at: http://robotics.ee.uwa.au/eyebot/. This includes the operating system used in the controller, as well as C and C++ compilers to run under Windows or Linux, image processing tools, a simulation system and a large collection of sample programs. A large amount of teaching material including PowerPoint slides is available from the same source.

The title of the book is slightly ambiguous since it is not immediately obvious what is meant to be embedded in what. The intention is that a versatile controller is embedded in many different robots.

The various considerations are treated in remarkable detail. In the first part of the book there are descriptions of various programming tools, and details of the inner working of the operating system, including its means of achieving multitasking with appropriate synchronisation depending on Dijkstra “semaphores”. Sensors and actuators are treated, as well as the means of real‐time image processing and a facility for wireless communication.

In the second part, various aspects of the design of mobile robots are discussed, including the various means that can be used to propel vehicles controllably. One of these is an omni‐directional drive. Balancing robots, walking robots and flying robots are also treated, as well as a simulator program.

The most remarkable part of the book is the third part in which successful applications are described. They include maze exploration, discussed in the context of a “Micro‐Mouse” international competition, and map generation, in which the robot has to explore and record a previously unknown environment which, unlike the test mazes that have been used, does not in general have all obstacles in a rectangular grid. Still more dramatic is the application of the principles to robot football, which is organised internationally in a number of different leagues. The most advanced league of all is that in which it is planned that humanoid legged robots should compete, but this is ahead of current technology. EyeBots have performed well in the next most advanced option where each wheeled player uses only data from its own sensors.

In preparation for the two later chapters this part has reviews of necessary aspects of artificial neural nets, genetic algorithms, and genetic programming. One of the later chapters then treats behaviour‐based systems, in which sensors are linked to actuators more directly than through a model of the environment (Brooks, 1999). A robot using a neural net is described, and is able to learn to locate a ball and to drive towards it.

The succeeding chapter presents really impressive results in achieving legged locomotion, including the biped version. The control methods were evolved using genetic programming, once a suitable framework had been set‐up using splines to compute smooth trajectories. The number of trials needed for the evolution were such that they were performed using simulation rather than the physical robot, but effective control seems to have been achieved surprisingly readily, considering the usually‐assumed difficulty of the task.

Everything is presented in a pleasant chatty style, and with a remarkable amount of detail, so that, for example, the theory of splines is explained and there are even program listings to show how they were computed. The quality of production is high, with numerous illustrations, some in colour. There is a very great deal here that anyone concerned with autonomous robots will certainly want to peruse, irrespective of whether it is EyeCons or controllers of some other kind that are embedded.

Further Reading

Brooks, R.A. (1999), Cambrian Intelligence: The Early History of the New AI, (reviewed in Kybernetes, Vol. 29 No. 4, p. 529, 2000, MIT Press, Cambridge, MA.

Related articles