|
Engineering and computer science are core elements of mobile robotics,
obviously, but when questions of intelligent behavior arise, artificial intelligence,
cognitive science, psychology and philosophy offer hypotheses and answers. Analysis of
system components, for example through error calculations, statistical evaluations etc.
are the domain of mathematics, and regarding the analysis of whole systems physics
proposes explanations, for example through chaos theory.
Second, autonomous mobile robots are the closest approximation yet of intelligent agents,
the age-old dream. For centuries people have been interested in building machines that
mimic living beings. From mechanical animals, using clockwork, to the software and
physical agents of artificial life - the question of "what is life?" and can we
understand it has always motivated research.
Perception and action are tightly coupled in living beings. To see, animals perform
specific head and eye movements. To interact with the environment, they anticipate the
result of their actions and predict the behavior of other objects. They alter the
environment in order to communicate (so-called stigmergy) nest building in ants is an
example of this.
Because of this tight coupling between perception and action there is a strong argument
for investigating intelligent behavior by means of situated agents, i.e. mobile robots. In
order to investigate simulations of life and lifelike agents that interact intelligently
with their environment, we need to close the loop between perception and action, allowing
the agent to determine what it sees. Whether we will have autonomous robots that match
human intelligence within 50 years, or whether humans will even be obsolete by then (very
fuzzy statements, because the definitions of "intelligent" and
"obsolete" are not at all clear), as some writers predict, or whether we will
have to wait another 100 years for truly intelligent household robots, as others reply,
autonomous mobile robots offer a uniquely suited research platform for investigating
intelligent behavior.
Third, there are commercial applications of mobile robots. Transportation, surveillance,
inspection, cleaning or household robots are just some examples. However, autonomous
mobile robots have not yet made much impact upon industrial and domestic applications,
mainly due to the lack of robust, reliable and flexible navigation and behavior mechanisms
for autonomous mobile robots operating in unmodified, semi-structured environments.
Installing markers such as beacons, visual patterns or induction loops (guiding wires
buried in the ground) is one way round this problem, but it is expensive, inflexible and
sometimes outright impossible. The alternative navigation in unmodified environments -
requires sophisticated sensor signal processing techniques which are still in their
experimental evaluation phases. Case studies in this book present some of these
techniques. So, to let mobile robots work in areas which are inaccessible to humans, or to
perform repetitive, difficult or dangerous tasks, is yet another strong motivation for
developing intelligent, autonomous robots.
And finally, there is also an aesthetic and artistic element to mobile robotics. Swarms of
robots collaborating to achieve a particular task, or moving about avoiding collisions
with one another and objects in their environment, beautifully designed mobile robots,
like for instance micro-robots, or miniature legged robots, appeal to our sense of
aesthetics. It is not surprising that mobile robots and robot arms have been used for
artistic performances.
Construct Your Own Working Robot
Mobile robotics, by nature, has to be practiced. There are a range of relatively cheap
mobile robots available now, which can be used for student practicals, student projects,
or robotics projects at home (robotics as a hobby is rapidly gaining ground). GRASMOOR,
built at the University of Manchester, is one example - it has its own on- board
controller, infrared sensors, light sensors, tactile sensors, and a differential drive
system'. GRASMOOR is controlled by a variant of the MIT 6270 controller, a controller with
analogue and digital inputs for sensors, and pulse- width-modulated output to drive motors
(the different types of sensors that can be used on robots are discussed in chapter 3, and
pulse width modulation generates electric pulses of variable length to drive motors at
variable speed). Like many robot micro-controllers, the 6.270 controller is based on the
Motorola 6811 microprocessor.
It is not difficult to get going for a few hundred dollars, using robot kits or technical
construction kits based on children's toys, some of which have micro-controllers and the
necessary software environment to program the robots. A good introduction to building your
own robot is Mobile
Robots : Inspiration to Implementation.
If you are competent at building electronic circuits - and they needn't be very
complicated - you can also use commercially available micro-controllers, and interface
sensors and motors to them to build your robot. The basic message is: you don't have to
invest large sums to build a mobile robot.
Experiments with Mobile Robots
This book contains 12 detailed case studies that cover the areas of robot learning,
navigation and simulation. Furthermore, there are examples, exercises and pointers to open
questions. One of their purposes is to indicate interesting areas of robotics research,
identifying open questions and relevant problems.
A fascinating introduction to thought experiments with robots is Valentino Braitenberg's
book on "synthetic psychology" (Vehicles),
which contains many experiments that can be implemented and carried out on real robots.
Organization of the Book
Scientific progress rests on the successes and failures of the past, and is only achieved
if the history of a scientific area is understood. This book therefore begins by looking
at the history of autonomous mobile robotics research, discussing early examples and their
contributions towards our understanding of the complex interaction between robots, the
world they operate in, and the tasks they are trying to achieve.
A robot, obviously, is made from hardware, and the functionality of a robot's sensors and
actuators influences its behavior greatly. The second chapter of the book, therefore,
looks at hardware issues specifically and discusses the most common robot sensors and
actuators.
A truly intelligent robot needs to be able to deal with uncertain, ambiguous,
contradictory and noisy data. It needs to learn through its own interaction with the
world, being able to assess events with respect to the goal it is trying to achieve, and
to alter its behavior if necessary. Chapter 4 presents mechanisms that can support these
fundamental learning competencies.
Mobility is (almost) pointless without the ability of goal-directed motion, i.e.
navigation. This book will therefore cover the area of mobile robot navigation, taking
some inspiration from the most successful navigators on earth: living beings (chapter 5).
Five case studies highlight the mechanisms used in successful robot navigation systems:
self-organization, emergent functionality and autonomous mapping of the environment
"as the robot perceives it".
Scientific research is not only about matter, it is about method as well. Given the
complexity of robot-environment interaction, given the sensitivity of a robot's sensors to
slight changes in the environment, to color and surface structure of objects, etc., to
date the proof of a robot control program is still in physical experiments. To know what
robot behavior will result from a specific robot control program, one actually has to run
the program on a real robot. Numerical models of the complex interaction between robot and
environment interaction are still imprecise approximations, due to the sensitivity of
robot sensors to variations in environmental conditions. However, chapter 6 looks at one
approach to construct a more faithful model of robot-environment interaction, and at the
conditions under which such modeling is achievable.
The purpose of this book is not only to give an introduction to the construction of mobile
robots and the design of intelligent controllers, but also to demonstrate methods of
evaluation of autonomous mobile robots - the science of mobile robotics. Scientific method
involves the analysis of existing knowledge, identification of open questions, the design
of an appropriate experimental procedure to investigate the question, and the analysis of
the results.
In established natural sciences this procedure has been refined over decades and is now
well understood, but in the relatively young science of robotics this is not the case.
There are no universally agreed procedures yet, neither for conducting experiments, nor
for the interpretation of results. Environments, robots and their tasks cannot yet be
described in unambiguous ways that allow independent replication of experiments and
independent verification of results. Instead, qualitative descriptions of experiments and
results have to be used. Widely accepted standard benchmark tests in the area of mobile
robotics do not exist, and existence proofs, i.e. the implementation of one particular
algorithm on one particular robot, operating in one particular environment, are the norm.
To develop a science of autonomous mobile robotics, quantitative descriptions of robots,
tasks and environments are needed, and independent replication and verification of
experiments has to become the standard procedure within the scientific community.
Existence proofs alone will not suffice to investigate mobile robots systematically. They
serve a purpose in the early stages of the emergence of a scientific field, but have to be
supplemented later by rigorous and quantitatively defined experimentation. Chapter 7
therefore discusses mathematical tools that allow such quantitative assessment of robot
performance, and gives three case studies of quantitative analysis of mobile robot
behavior.
The book concludes with an analysis of the reasons for successes in mobile robotics
research, and identifies technological, control and methodological challenges that lie
ahead.
Mobile robotics is a vast research area, with many more facets than this introductory
textbook can cover. The purpose of this book is to whet your appetite for mobile robotics
research. Each chapter of this book includes pointers to further reading, and world-wide
web links, in addition to the references given in the text. Using these, you will
hopefully agree that mobile robotics is indeed a fascinating research topic that throws
some light on the age-old question:
"What are the fundamental building blocks of intelligent behavior?"
|
