Page 15 - Innovations in Intelligent Machines

2 L.C. Jain et al.
When we talk about intelligent machines, the first thing that normally
appears in our mind is robots. Indeed, robots have been invented to substi-
tute humans in performing a lot of tasks involving repetitive and laborious
functions, for examples pick-and-place operations in manufacturing plants.
However, robots that are operated based on a programmed manner and in a
fully controlled environment are not considered as intelligent machines. Such
robots will easily fail when the application and/or the environment contain
some uncertain condition. As an example, in applications that involve haz-
ardous and uncertain environments such as handling of radioactive and explo-
sive materials, exploration into space and ocean, robots that can react to
changes in their surrounding are very much needed. As a result, robots have
to be equipped with “intelligence” so that they can be more useful and usable
when operating in uncertain environments.
To be considered as an intelligent machine, the machine has to be able to
interact with its environment autonomously. Interacting with the environment
involves both learning from it and adapting to its changes. This characteristic
differentiates normal machines from intelligent ones. In other words, a normal
machine has a specific programmed set of tasks in which it will execute accord-
ingly. On the other hand, an intelligent machine has a goal to achieve, and it
is equipped with a learning mechanism to help realize the desired goal [3].
The organization of this chapter is as follows. In section 2, some learning
methodologies for intelligent machines are discussed. In section 3, applications
of intelligent machines to a number of areas including unmanned aerial vehi-
cles, robots for space and ocean exploration, humanoid robots are presented.
A description of each chapter included in this book is presented in section 4,
and a summary of this chapter is included in section 5.

2 Learning in Intelligent Machines

When tackling learning from the machine perspective, Artificial Intelligence
(AI) has become one of the main fields of interest. The definition of AI can be
considered from three viewpoints [4]: (i) computational psychology–mimicking
and understanding human intelligence by the generation of a computer
program that behaves in the same way; (ii) computational philosophy–
formulating a model that is implementable in a computer for understanding
intelligent behaviors at the human level; and (iii) machine intelligence–
attempting to program a computer to carry out tasks, until recently, only
people could do.
In general, the learning process in intelligent machines involves acquiring
information about its environment, and deploying the information to establish
knowledge about the environment, and, subsequently, generalizing the knowl-
edge base so that it can handle uncertainty in the environment. A number of
machine intelligence techniques have been developed to introduce learning in
machines, e.g. imitation learning [5] and reinforcement learning [6]. For robot