Название: The Developmental Organization of Robot Behavior
Автор: Roderic Grupen
Издательство: The MIT Press
Серия: Intelligent robotics and autonomous agents series
Год: 2023
Страниц: 389
Язык: английский
Формат: epub (true)
Размер: 34.8 MB

A comprehensive introduction to the mathematical foundations of movement and actuation that apply equally to animals and machines.

This textbook offers a computational framework for the sensorimotor stage of development as applied to robotics. Much work in developmental robotics is based on ad hoc examples, without a full computational basis. This book's comprehensive and complete treatment fills the gap, drawing on the principal mechanisms of development in the first year of life to introduce what is essentially an operating system for developing robots. The goal is to apply principles of development to robot systems that not only achieve new levels of performance but also provide evidence for scientific theories of human development.

Robotics is a unique discipline in that it affords us the ability to study sophisticated machines and computation simultaneously in a platform where we can precisely control experimental studies of intelligence. The discussion in the book begins by studying the embodied systems on which development depends—in particular, how embodied structure can be used to compensate for the lack of structure in the open environment. The discussion attempts, wherever possible, to compare and contrast human bodies with their robotic counterparts. We explore the mathematical foundations of sensory and motor systems, that might normally be covered in a first course in robotics (kinematics, dynamics, actuators, sensors, control, and signal processing). Much of the supporting mathematics is introduced in the appendices so as not to interfere with the flow of the book. The more advanced (and specialized) material presented (knowledge representations, biological sensors, muscles and mechanisms, developmental neurology, potential fields, learning, and redundant systems) is critical to the overall thesis of the book. It provides much of the representational framework that can be used to model development in robot systems.

The text has been used in an undergraduate course entitled Introduction to Robotics, which caters to computer scientists and engineers as an introduction to the mathematics underlying robot systems, perception, and control. The content presented in chapters 2, 3, 4, and 6 represent the core material for many undergraduate introductions to robotics. A recent offering at the University of Massachusetts used chapters 1 (introduction), 2 (actuation), 3 (control), 4 (kinematics), 6 (dynamics), and 8 (signal Processing), with selected subjects from other chapters and eight written and related programming assignments. The complete text (with liberal use of the written and programming assignments) has been used successfully in a graduate course (primarily Computer Science and engineering students) that focused on the special properties of embodied Machine Learning systems.