Page 25 - Toucpad robotics C11

A task can be created where the model has to predict a word that has been intentionally masked out in a sentence.
For instance, in the sentence “The capital of France is [MASK],” the model learns to predict ‘Paris’. By solving millions
of such ‘fill-in-the-blank’ puzzles, the model learns the grammar, vocabulary, and meaning of the language, even
without anyone manually labelling the text. This foundational understanding can then be used for more specific tasks
like translating a sentence or summarising a long article.
Applications:
u
Object Recognition and Classification: Identifying different types of items on a conveyor belt.
l
Predictive Maintenance: Learning from historical data to predict when a robot’s component might fail, allowing
l
for proactive repairs.
Robotic Grasping: Teaching a robot how to pick up objects of varying shapes and sizes.
l
Autonomous Navigation: Learning optimal paths and behaviours in complex environments.
l
Example: Imagine training a robot to learn how to walk. Using Reinforcement Learning, the robot might try different
u
leg movements. If it stays balanced and moves forward, it receives a ‘reward’. If it falls, it receives a ‘penalty’. Over
many trials, it ‘learns’ the most effective sequence of movements to walk steadily.
Computer Vision
Concept: This is a field of AI that enables robots and computers to “see” and interpret visual information from the
u
real world, just like humans do. It involves processing images and videos to extract meaningful information.
Applications:
u
Object Detection and Tracking: Identifying specific objects (e.g., a person, a car, a traffic light) and following
l
their movement.
Scene Understanding: Interpreting the overall context of a visual scene (e.g., “this is a busy street,” “this is a
l
quiet office”).
Visual Servoing: Using camera feedback to precisely control robot movements (e.g., aligning a part for assembly).
l
Quality Inspection: Automated visual inspection of products for defects.
l
Example: A self-driving car heavily relies on computer vision. Its cameras capture real-time video of the road.
u
AI algorithms then process these images to identify traffic lights (red, yellow, green), road signs (stop, turn), lane
markings, pedestrians, and other vehicles, enabling the car to make informed driving decisions.

Natural Language Processing (NLP)
Concept: NLP is an AI field that deals with the interaction between computers and human language. It allows robots
u
to understand, interpret, and generate human language (both spoken and written).
Applications:
u
Voice Control: Enabling robots to understand and respond to spoken commands.
l
Human-Robot Interaction: Facilitating more natural and intuitive communication between humans and robots.
l
Chatbots and Virtual Assistants: Though not physical robots, these demonstrate NLP capabilities.
l
The machine responds A human talks to the
with an audio file machine

Data-to-audio conversion How NLP The machine captures
occurs Work the audio

The machine processes Audio-to-text
the text's data conversion takes place

23
Introduction to Robots: What Exactly are They?