Associative Learning for a Robot Intelligence

Citation

Rigelsford, J. (2001), "Associative Learning for a Robot Intelligence", Industrial Robot, Vol. 28 No. 4. https://doi.org/10.1108/ir.2001.04928dae.003

Associative Learning for a Robot Intelligence

J.H. AndreaeImperial College Press1998348 pp.ISBN 1-86094-132-X£35.00 (Hardback)

Keywords: Robots, Intelligence, Learning

This book comprises 13 chapters and aims to attract new researchers to the unexplored potential of associative learning. Chapter 1, associative learning, introduces the concepts of a robot's learning brain, the fundamental limits of associative learning, compromises, and justification from a finite state machine.

The BunPie microworld is presented in chapter 2 and discusses the concepts of motivation, reward, plans and looking ahead. Chapters 4 and 5 discuss numbers in the head, and the Universal Turing Machine (UTM), respectively. Subjects discussed include learning to count, streams and templates for a UTM, teaching the counting, and the quintuples of the UTM.

Chapter 6, communicating intentions, discusses boredom and frustration, the first moment experiment, making a plan, and asks the question, "What does a plan mean?". The concept of consciousness before language is discussed in chapter 7, while a hierarchical task is presented in chapter 8. The following two chapters address painted vision and co-operation, respectively. Topics covered include assessing painted vision, the emergence of co-operation, planning in the co-operation cycle, and the wealth of stimulus.

Chapter 12, turn taking, discusses five-fingered hands, learning different templates and imitation. The final chapter of the book, climbing a tree or building a rocket, gives further discussion of PRR-PUSS – a vehicle for exploring associative learning, and an engineer's dream, the concluding remarks of the author.

Associative Learning for a Robot Intelligence is more of a philosophical text than an engineering one. Although the algorithms are intended for real-world robots, unfortunately there is no sign of a robotic test bed, and the presentation of procedures could easily be dismissed as of little practical use.

Overall, this book is definitely different but unfortunately failed to convert me to be a disciple of associative learning.

Jon Rigelsford