Glossary
Robotics and AI glossary
Short, plain-English definitions of the terms you will meet while learning.
A
- Activation function
- A function applied to a neuron's output that lets a neural network learn non-linear patterns. Read: How neural networks learn
- Actuator
- A component, such as a motor or pneumatic cylinder, that converts energy into motion or force. Read: Sensors and actuators
B
- Backpropagation
- The algorithm that calculates how much each weight in a neural network contributed to the error so the weights can be updated. Read: How neural networks learn
- Bias (model)
- Systematic error in a model's output, often caused by unrepresentative or skewed training data.
- Bounding box
- A rectangle drawn around an object in an image to mark its position, used in object detection. Read: Computer vision for robots
C
- Closed-loop control
- Control that measures the result of an action with a sensor and adjusts the command to reduce error. Read: Control loops and PID
- Computer vision
- The field of enabling computers to extract information from images and video. Read: Computer vision for robots
- Convolutional neural network (CNN)
- A neural network designed for images that detects local patterns such as edges and shapes using small sliding filters.
D
- Deep learning
- Machine learning that uses neural networks with many layers. Read: AI vs ML vs deep learning
- Degrees of freedom (DOF)
- The number of independent ways a robot or joint can move. A simple hinge joint has one.
E
- Encoder
- A sensor that measures the rotation of a shaft or wheel, used to track speed and position. Read: Sensors and actuators
- End effector
- The tool at the end of a robot arm, such as a gripper or welding head.
- Epoch
- One complete pass through the entire training dataset during model training.
G
- Gradient descent
- An optimisation method that repeatedly adjusts parameters in the direction that reduces the loss. Read: How neural networks learn
I
- IMU (inertial measurement unit)
- A sensor package, usually an accelerometer and gyroscope, that measures acceleration and rotation rate.
- Inverse kinematics
- Calculating the joint angles a robot arm needs so that its end effector reaches a desired position.
L
- LiDAR
- A sensor that measures distance by timing reflected laser pulses, producing a detailed map of surroundings.
- Loss function
- A measure of how wrong a model's predictions are, which training tries to minimise.
M
- Machine learning
- A branch of AI in which systems learn patterns from data instead of following only explicit rules. Read: What is artificial intelligence?
- Microcontroller
- A small, low-power computer on a single chip used to read sensors and drive actuators, such as the chip on an Arduino board.
O
- Occupancy grid
- A map made of cells, each marked as free, occupied or unknown, used for robot navigation. Read: Path planning and SLAM
- Odometry
- Estimating a robot's change in position from motion data such as wheel rotation. It drifts over time.
- Overfitting
- When a model memorises its training data and performs poorly on new data. Read: How neural networks learn
P
- PID controller
- A feedback controller that combines proportional, integral and derivative terms of the error to produce its output. Read: Control loops and PID
- PWM (pulse-width modulation)
- Switching a signal on and off rapidly and varying the on-time to control motor speed or LED brightness.
R
- Reinforcement learning
- Learning by trial and error, where an agent receives rewards for desirable actions. Read: What is artificial intelligence?
- ROS (Robot Operating System)
- An open-source framework of tools and libraries for building robot software, including navigation and SLAM packages.
S
- Sensor fusion
- Combining data from multiple sensors to obtain a more accurate or reliable estimate than any one sensor gives.
- Servo motor
- A motor with built-in feedback that moves to and holds a commanded angle. Read: Sensors and actuators
- SLAM
- Simultaneous localisation and mapping: building a map of an unknown place while tracking position within it. Read: Path planning and SLAM
- Stepper motor
- A motor that rotates in small fixed steps, allowing accurate positioning without feedback sensors.
- Supervised learning
- Machine learning from examples that include the correct answers (labels). Read: What is artificial intelligence?
T
- Torque
- The rotational force a motor can apply. Gearing trades speed for torque.
- Transformer
- A neural network architecture built on attention, widely used for language and other sequence data.
U
- Unsupervised learning
- Machine learning that finds structure in data without labelled answers.
W
- Weight
- A learned number in a neural network that scales an input's influence on a neuron. Read: How neural networks learn
Frequently asked questions
What is the difference between a sensor and an actuator?
A sensor measures something such as distance or rotation and sends data to a controller. An actuator converts a command into motion or force, as a motor does.
What does SLAM stand for?
SLAM stands for simultaneous localisation and mapping: building a map of an unknown place while tracking your own position within it.
What is the difference between machine learning and deep learning?
Machine learning is any approach where a system learns patterns from data. Deep learning is a type of machine learning that uses neural networks with many layers.
