Page 23 - Toucpad robotics C11

The Power of Artificial Intelligence (AI) in Robotics

While sensors give robots their ‘senses’ and mechanical
structures give them their ‘body’, it is Artificial Intelligence
(AI) that provides the ‘intelligence’ – the ability to reason, learn,
make decisions, and adapt. AI is what transforms a simple
machine into a truly intelligent system.

What is AI in Robotics?
In robotics, AI is the field that equips robots with cognitive
abilities, allowing them to:
Perceive: Interpret complex sensor data (e.g., understanding
u
what a camera image shows).
Reason: Make logical deductions and plan actions based on its understanding.
u
Learn: Improve its performance over time through experience or data.
u
Decide: Choose the best course of action in uncertain or changing environments.
u
Essentially, AI helps robots to think and behave in ways that would typically require human intelligence.

Key AI Techniques Used in Robotics
Several branches of AI are particularly vital for robotics:

Supervised Learning
Imagine you are preparing for a Mathematics exam. Your teacher gives you a textbook filled with solved problems. You
study these problems, understanding the steps and solutions for each type. When you face a new problem in the exam,
you can apply what you’ve learned from the solved examples to find the correct answer.
This is exactly how Supervised Learning works. The “solved problems” are the labelled data, where the inputs
(the problems) are linked to the outputs (the solutions). The machine learning model is like you, the student, learning to
map the inputs to the correct outputs. It learns from these examples so that when it sees new, unseen data, it can make
accurate predictions or classifications.
Example: Think of a spam filter in your email. When you mark an email as ‘spam,’ you are essentially giving a
u
label to it. The system learns from thousands of these labelled emails—some marked as ‘spam’ and others as ‘not
spam.’ Over time, it learns to identify characteristics of spam emails (like certain keywords or sender addresses) and
automatically sends new, incoming spam emails to your spam folder without you having to manually label them.

Unsupervised Learning
Now, imagine your teacher gives you a large, unsorted pile of books from the school library. They don’t tell you which
subject each book belongs to. Your task is to organise them on the shelves. You start by looking at the books and notice
that some are about science, others are about history, and a few are storybooks. You group the similar-looking books
together, even without a label, purely by observing their characteristics—like the content, pictures, or title.
Unsupervised Learning is just like this. The machine learning model is given a large dataset with no labels or guidance.
Its job is to find hidden patterns, group similar data points together (this is called clustering), and discover structures on
its own. It’s about finding order in chaos.

u Example: A popular streaming platform like Netflix or Amazon Prime Video uses unsupervised learning to recommend
movies. They analyse the viewing habits of millions of users—what movies they watch, how long they watch, and what
they rate highly. The system then groups users with similar tastes together and recommends new movies to them
based on what others in their group have watched and enjoyed. The system learns these patterns without ever being
told, “this person likes action movies.”

21
Introduction to Robots: What Exactly are They?