We've been awed by the depiction of robots with intelligence equaling humans
in science fiction movies, and in novels which talk about the exciting future of
robots and androids after the turn of the century. Robots that would be doing
tasks ranging from daily household chores to complex jobs like flying a
spacecraft and having the ability to take decisive actions in critical
circumstances. Now, when we are 8 years past the turn of century, such science
fiction robots with artificial intelligence (AI) are yet to become a reality.
This doesn't mean that work and research aren't taking place in field of
robotics and AI. Research work and technological advancements are happening in
these domains at a steady pace and future of such intelligent robots working at
par with humans is not distant. In this article, we'll discuss some new concepts
and research work taking place in robotics and AI.
Robotics meets AI
Robotics, started in 1950s as a result of an automation project, has made
tremendous technological advancements. It has become a part of our life, and we
have become technologically addicted to its use. Today, in many industries
robotics have replaced human workers, where they are used for heavy forklift
duties and warehouse housekeeping. Robotic machines have spared humans from jobs
that were both repetitive and hazardous. A classic example of robotics being
used in an industry widely would be of automobile industry, where jobs like
welding to final paint of the finished product is done with robotic machines and
with minimal human help. Even fields like medicine, military, etc, are no
strangers to Robotics. In recent times, we've seen robotic arms being used
during surgeries where precision is a must. In military, robots are used for
clearing minefields; fire-fighting departments use robotic vehicles to venture
into parts which are out of human reach. These machines are mostly mechanical,
doing tasks that were either continuous or repetitive, or involving precision,
or were hazardous for humans to perform.
Robots are now evolving from industry machines to take form of humans or even
animals. With use of techniques from AI, robots like ASIMO which are of human
form, are being programmed to perform tasks like ascending and descending
stairs; a task that requires maintaining balance with center of gravity. AI,
when defined in simple terms, is a science of engineering intelligent machines
that will possess traits to exhibit power for reasoning, knowledge, planning,
learning, communication, perception, and the ability to move and manipulate
objects. This sort of power to reason and to decide is mostly done unconsciously
by a human mind, but for a machine to decide for a path or to do a task, it will
have to be programmed for a set of rules and models. Following such set of rules
and determining the best possible result can best be exampled for various chess
moves.
IBM's Deep Blue computer was the first AI-based machine to defeat a human in chess and that too the then reigning world champion Gary Kasparov. |
Conventional robotics is largely based on having robots that carry out
complex calculations, for instance measuring the geometry of an object and its
trajectory when it's in motion. A human child, who is unaware of the
fundamentals of the geometry, can still figure out how to catch a ball mainly
because the child can learn from experience and observation. The same process of
self-learning from experience and observation can't be expected from robots as
they follow pre-defined rule-based evaluation process. Making a robot follow a
decision-making process like climbing branches of a tree would be based on
approach of AI where a rule-based system supplies the knowledge and scene
representations. And based on knowledge and determining the scenes, the robot
will be able to determine and make a decision which branches to hold on to climb
the tree.
There are many approaches and methodologies in AI that are being implemented
into robotics so that a robot can exhibit intelligence to recognize, learn,
adapt, and decide to react and function on its own for situational tasks. The
ability to learn from trial and error and with experience will make a robot to
make more independent decisions rather than a decision based on pre-defined set
of rules. Across the globe there is some exciting research taking place to make
AI as capable as a human mind.
Artificial Neural Networks
Now biologically inspired artificial neural networks (ANNs) based robots are
being developed that will not be relying on preset rules, but on continuous
processing of signals and non-linear optimization process to reach a response to
act autonomously. Neural Network is basically an information-processing
architecture that is designed to determine and track relationships among various
data sets autonomously. It's a system constructed along the same principles as a
human brain that performs mathematical and analogical calculations between one
domain or layer to another using algorithmic techniques. What sets ANNs apart
from classical AI methods is that a neural network includes a 'self-learning'
mechanism that allows the ANN-based systems to analyze almost huge amount of
data, do test on them and determine relationships among the data, and then use
the findings of the analysis with its preset rules and formula to make a
prediction about the event. Additionally, the ANN-based system compares its
prediction to actual results to learn over time, much like a human brain does,
by adjusting the preset formula to reduce discrepancies between the predicted
and the actual results. A system or a robot having ANN-based intelligence will
be able to produce more accurate results and predictions over time, as it will
be learning from its mistakes and will be avoiding them in future by adjusting
the formula and rules. This ability of ANNs brings the much-desired learning
capability into a robot to make its own decisions by discovering relationships
and correlating them. The advantage of using ANNs with classical AI in robotics
leads to performance of tasks that humans usually do sub-consciously, for
instance locomotion or cognitive tasks. Such intelligence systems can be of
great value not only in robotics but also in real life scenarios. For instance,
making market trend predictions in financial analysis, or working as an
auto-pilot mechanism in an aircraft thereby increasing safety.
Robot 'Kismet' recognizing or reacting to human gestures, based on its emotional intelligence program. |
Emotional intelligence
In the movie 'I,Robot', the robot Sonny was depicted to feel emotions though
it was just a machine having AI. Researchers have been working at the
possibilities where even robots will be able to react to emotions and also show
emotions. For this purpose, a non-symbolic emotion model was developed that
takes the form of a recurrent artificial neural network where emotions both
depend on and influence the perception of the state of the world. This
emotion-based model was integrated to a self-learning AI architecture to
determine influence from perceptions and to evaluate the decisions based on
that. The research work showed that the artificial emotions are useful when
incorporated into classical AI to have behavior-based systems or robots that
will be able to play a part of the society giving an impression of empathy and
serving humans by helping sick and elderly at hospitals or people who are
house-bound and help them as domestic helpers who would be in tune to their
owners emotions. Emotional intelligence has been achieved to a certain degree in
form of smile, frown, and reactions to tones. Based on voice pitch an emotion
reading system will decipher if it's harsh, frustration or a simple voice
command. Sony's AIBO is one such AI based robot that has been introduced as
robotic pets; these pets can learn from human stimuli and also react to such
stimuli like they portray happiness when being played with, whereas they whimper
when put in seclusion. This is done by the cognitive intelligence where the
human facial recognition and voice commands can be interpreted by emotional
intelligence system to trigger stimulus in the robot, AIBO.
NanoRobots
Nanorobots, robots of microscopic scale of about nanometers, are based on
hypothetical nanotechnology engineering to build machines that are microscopic
in size. Building robots that would be so small that many of them could be lined
across the width of a human hair is a big challenge to achieve. Already we have
seen that microprocessor chips are being manufactured in microscopic sizes, and
so we can expect that a robot having such a chip can do predefined tasks. The
challenge is to build a machine of such scale that wraps around the chips. On an
own, a single nanorobot will not be able to function much, but in numbers they
can work together to do a task that requires high precision and is of complex
nature. The usage of such tiny robots, nanorobots, can be aplenty and mainly in
field of medicine. Where nanorobots will be injected into the blood system of a
human to treat at the cellular level, for instance a tumor can be dissolved.
Same way, during surgeries, nanorobots can be injected into the blood stream
where they will form a blockage at the arteries so as to prevent excessive blood
loss. Though there hasn't been any nanorobot manufactured yet, but in coming
years we can see this technological barrier will also be overcome.