The chapters in this book are logically selected and grouped.
The path that is followed goes through four stages:
* Research inspired by biological systems at the behavioral level
* Control of robot arms using artificial neural networks
* Simulation of and inspiration by biological neural systems
* Control and navigation of mobile robots using artificial
The first three chapters describe neural networks which simulate
biological systems at the behavioral level. The third chapter ends
with neural control of a robot arm; this topic is picked up by the
subsequent---overview---chapter, followed by an in-depth study in this
field. The next three chapters are focused on biological neural
systems, and describe applications in the navigation of mobile robots.
This theme is covered in detail in the final two chapters.
Evaluating a biological system at the behavioral level, Chapter 1,
``Neural Network Sonar as a Perceptual Modality for Robotics,'' by
Itiel Dror, Mark Zagaeski, Damien Rios, and Cynthia Moss, describes a
neural network which approximates echo-locating behavior of the big
brown bat, Eptesicus fuscus. Using previous studies of this bat, a
neural system is introduced which can determine speed of movement using
a single echolocation only, referring back to studies which show that
bats differentiate between different wingbeat rates of insects. The
results presented in this chapter provide a good basis for the use of
echolocation in robotic systems.
In Chapter 2, ``Dynamic Balance of a Biped Walking Robot,'' by Thomas
Miller III and Andrew Kun, a neural system is used to have a robot
learn to walk. The approach is unique: Instead of using analyses of
walking behavior of biological systems, the neural network-driven robot
uses feedback from force sensors mounted on the undersides of the feet,
as well as from accelerometers mounted on the body. The learning
behavior that is exhibited typically resembles that of biological
systems which learn to walk.
A technique for the control of robot manipulators is introduced in
Chapter 3, ``Visual Feedback in Motion,'' by Patrick van der Smagt and
Frans Groen. This research is also inspired by a biological system at
the behavioral level. Using studies of the gannet from the family of
Sulidae, sequences of two-dimensional visual signals are interpreted to
guide a monocular robot arm in three-dimensional space without using
models of the visual sensor nor the robot arm.
Exploration of the control of robot arms is continued in Chapter 4,
``Inverse Kinematics of Dextrous Manipulators,'' by David DeMers and
Kenneth Kreutz-Delgado. The chapter gives an overview of neural and
non-neural methods to solve the inverse kinematics problem: Given an
end-effector position and orientation, how should one move a robot arm
(in a most efficient way) to reach that position/orientation?
The theoretically inclined Chapter 5, ``Stable Manipulator Trajectory
Control Using Neural Networks,'' by Yichuang Jin, Tony Pipe, and Alan
Winfield, describes neural network approaches for trajectory following
of a robot arm. The key issue here is how to improve the accuracy of
the followed trajectory when the dynamic model of the robot arm is
Studies of sensory motor control in biological organisms and robots are
presented in Chapter 6, ``The Neural Dynamics Approach to Sensory-Motor
Control,'' by Paolo Gaudiano, Frank Guenther, and Eduardo Zalama. It
extensively discusses neural network models developed at Boston
University's Center for Adaptive Systems. The neural models are used
in two applications: trajectory following of a mobile robot, and
controlling the motor skills required for speech reproduction using
Biomorphic robots are discussed in Chapter 7, ``Operant Conditioning in
Robots,'' by Andreas B\"uhlmeier and Gerhard Manteuffel. In their
overview chapter, they discuss neural systems which maintain
homeostasis for (mobile) robot systems. After discussing neural
learning systems with neurophysiological backgrounds, a survey of
several implementations on mobile robots, which have to learn to
navigate between obstacles, is given.
In Chapter 8, ``A Dynamic Net for Robot Control,'' by Bridget Hallam,
John Hallam, and Gillian Hayes, a neural model, designed for explaining
various learning phenomena from animal literature, is used to control a
The navigation of mobile robots using artificial neural networks is
covered in Chapter 9, ``Neural Vehicles,'' by Ben Kr\"ose and Joris van
Dam. The authors make the distinction between reactive navigation,
planned navigation in known environments, and map building from sensor
In the final chapter, ``Self-Organization and Autonomous Robots,''
Jukka Heikkonen and Pasi Koikkalainen describe the use of
self-organizing maps for reactive control of mobile robots.
dr Patrick van der Smagt phone +49 8153 281152
DLR/Institute of Robotics and Systems Dynamics fax +49 8153 281134
P.O. Box 1116, 82230 Wessling, Germany email <smagt at dlr.de>