Robotics Dictionary

by on August 1, 2022

This is functional dictionary for roboticists, and not a proper dictionary. The goal of making this is to define terms from within the research and engineering world in an easy to understand manner.

Try to read through this list to test your robotics terminology.

This is a work in progress and I will be continuing to add to this list. Please leave a comment below if you have a word/term that I should add.

All0-9ABCDEFGHIKLMOPQRSTUV
0-9
80/20 (8020)
T slot aluminum extrusion with many joints and accessories. Very useful for building testbeds or other structures. Almost like lego for adults. 80/20 is a specific company but is often the name people use for similiar products made from other companies.
A
A*
Search algorithm for known maps for robot navigation. Also used for searching data structures and other computer science applications.
Adaptive Data Augmentation
The idea of modifying (scaling, adding noise, rotating, etc…) existing training data so that you can use it as “new” data for further training a model.
After Action Report (AAR)
A report or summary of what happened after an event, where the event is often a test of the robot. This differs from a hot wash which should occur immediately after an event, with that said doing an AAR quickly will help you remember what happened better..
Axis
1. Refers to a degree of freedom available on the robot.
2. In motion control this refers directly to each motor.
B
BVLOS
Beyond Visual Line of Sight (Usually for drone operations, but can be used for other remote operations)
Behavior Tree (BT)
A flow chart (tree) that controls how a decision is reached within a system.
Behaviors
Plans for what a robot should do (or how to behave) in a given situation.
C
Closed Loop Control
Using feedback to verify that a command occurred (and adjusting based on feedback). For example you command a robot to spin and use an encoder to control commands to make sure it happens as desired.
Configuration Space (c space)
The space where the robot can operate in. Think of a robot arm, where are the spaces that it can reach?
Covariance Matrix
A matrix of Covariance’s. Covariance is a measure of variability between two values. Often used for expressing variance for error terms.
Critical Design Review (CDR)
A review of a design after the majority of the detailed design is complete and before ordering and fabrication begins. This review can include the technology, the schedule, the cost, or the ability to meet any other requirement.
D
D*
Dynamic A* Search. Search algorithm based on A* that allows for fast replanning in a dynamic world.
DIN Rail Terminals
DIN is a standard rail system that you can connect devices to and is often used for power distribution. Devices can include terminal blocks, fuses, and even larger equipment such as network switches and computers. See here for some basic DIN hardware https://www.mcmaster.com/terminals/modular-din-rail-mount-terminal-blocks-11/
Degrees of Freedom (DoF)
An axis that can be moved/controlled by the robot.
Often uses as a question, How many degrees of freedom does a robot have?
Direct Drive
Directly driving a load or mechanism from a motor with no intermediate gearing. This is often nice for eliminating backlash or for more precise current/torque control.
Draw bar pull
The pulling power of a robot. If you attach a sled to the back, how much weight can you put in the sled?
Drift
Erroneous change in a value over time. Often in sensors used, such as in gyro’s (within in IMU). This can cause drift in localization or a robot or in maps generated by a robot.
Dynamic Movement Primitives (DMP)
Method of trajectory control and planning. DMP’s use a set of nonlinear predefined models/equations to specify a motion from point A to point B, instead of a specific predefined motion primitive library. 
E
E-Stop (estop)
Emergency Stop. A button or control that kills power to motors/robot
End Effector
Fancy word for the gripper/tool at the end of a robotic arm
Ergodic Search
A search method using an ergodic metric. An ergodic metric tries to maximize the time an agent (in the case of a multi-agent system) searches areas with high information, relative to areas with low information.

This is a method of exploration that covers all areas, but is weighted to spend more time searching known areas.
Euler angles 
See Roll Pitch Yaw (RPY). Yes, I know there is more to this topic…
F
Factor graph
A type of graphical model that has variables and equations to model those equations. Factor graphs are used for modeling and solving inference problems including for simultaneous localization and mapping (SLAM).
Female
The part of a connector (or socket) that the other end gets inserted into.
Forward Kinematics (FK)
Equation(s) to compute the position of the end-effector based on current values of the joints
Four-bar linkage (Four bar)
Common mechanical mechanism. Consists of four rigid bars connected by pins or planer joints. Four bar mechanisms can provide force multiplication.
Frontiers
 Regions on the map border between open/explored space and unexplored space when doing exploration.
G
GTSAM
C++ library that implements the factor graph data structure for robotics and computer vision applications, including SLAM.
H
Hot Wash
This is a term used for an immediate after action report for something that happened. Hot washes often occur right after a field test when the details are still “hot” in everyone’s mind.
I
ICAM
Interpretable Classification via Disentangled Representations and Feature Attribution Mapping. A machine learning framework for learning features from images.
Inertial Measurement Unit (IMU)
Sensor that combines gyros, accelerometers and/or magnetometers into a single package. Often has sensors to measure 3 axis for roll, pitch, and yaw. Often also can provide raw sensor data for computing motion and rotations on your own. If a magnetometer is provided the sensor can provide magnetic yaw angle (assuming no interference).
Inference time
How long it takes for a model to process incoming data and produce a result/prediction. We often want to know is the inference time good enough to run in “real time” as data is produced and we need the result.
Intersection over Union (IoU)
Metric to measure the overlap between two boxes. Often used in machine learning to compare a new image to a ground truth image. Range is generally 0-1, where 1 is exactly the same (which does not happen in practice). Everybody has different ideas of a good value, I will claim >0.7 is good.
Inverse Kinematics (IK)
Equation(s) to determine the motion of a robot to reach a desired position
Iterative closest point (ICP) 
Algorithm to minimize the distance between two point clouds. Often used for SLAM and calibration to line up point clouds from different sources/time.
K
Kalman Filter (KF, EKF)
Used for estimating state by estimating state then correcting the estimate over time. https://robotsforroboticists.com/kalman-filtering/
Kinematics
Method determining position of joints or how to get a given position. See Forward Kinematics and/or Inverse Kinematics.
L
Laser Odometry
Determining your motion based on laser sources (ie LIDAR).
Lidar Odometry and Mapping (LOAM)
Algorithm for real time simultaneous localization and mapping (SLAM).
Linear Quadratic (LQR) Controller
A method of control that allows for using non-linear (or linear) models to estimate in advance what controller commands will be needed to successfully execute a desired command.

This is often used instead of PID controllers that are just reactive to error in commands.
M
M-Stop (mstop)
Motion Stop. A button or control that halts motion (but does not kill power)
Male
The part of a connector (or pin) that goes into the receiving female end (or socket) side of the connector.
Model Predictive Path Integral (MPPI)
An iterative method for controlling trajectories. After a trajectory is optimized and started to be executed, the trajectory is re-optimized using the remaining un-executed portion of the trajectory.
Motion Primitives
Usually this is a set of predefined motions to get from point A to point B that a planner can select. This allows the planner to not worry about the details of the motion, just which motion is correct. In some cases, such as DMP the primitives can take the form of equations.
Motor Amplifier
The hardware that commutates and powers the motor. The motor amplifier is often combined with a motor controller into a common package.
Motor Controller
The hardware that is used to control a motor, such as to use feedback and run a PID loop (but not necessarily commutate the motor). This term is often used for a motor controller with an an integrated motor amplifier.
MuJoCo
Name of a physics simulator often used in robotics.
Multi Agent System
Using multiple robots at the same time
O
Open Motion Planning Library (OMPL)
OMPL consists of many sampling-based motion planning algorithms. It does not contain code related to other aspects needed in controls such as collision checking or visualization.
Open loop control
Controlling something and not observing if it is actually happening. For example sending a command to a motor and just hoping that it ends up where you want it.
P
Point Cloud Library (PCL)
 PCL is a standalone, large scale, open project for 2D/3D image and point cloud processing.
Preliminary Design Review (PDR)
A review of a design before the detailed design begins. This makes sure that the assumptions and overall technological approach is correct, the PDR will also usually look at schedule, budget and other requirements to make sure reviewers believe they can be met. This is often followed up with a Critical Design Review (CDR).
Proportional Integral Derivative (PID) Controller
Reactive method of control to minimize error while taking into account the error, the rate the error changes, and the error over time. Often only one or two of the terms are used in a controller. See here for more details https://robotsforroboticists.com/pid-control/
Q
Quaternion
A fancy way of expressing roll, pitch, and yaw (RPY). In certain applications (such as in space) it can prevent some issues that are possible with RPY notation. RPY is easier for humans to interpret, however quaternions are nice for computation.
R
RGBD
Generally refers to a camera that has Red, Green, Blue and depth channels.
Rapidly-exploring Random Tree (RRT)
Randomly builds a tree in unknown space to plan a path through a space for navigation
Recovery Behaviors
What a robot should do (or how to behave) if stuck or in a bad situation.
Red’s Rule
Red’s Rule states that a single wheel (ie motor) on a wheeled robot should have enough torque to pull the entire robot.
Reinforcement Learning
A type of machine learning where a system learns from its mistakes. Uses trial and error with appropriate feedback to know if an action was correct.
Robot
I define a robot as something that senses, processes the data, and acts.
So yes, some fancy new dishwashers can be a robot.
Robustify
To make robust
Roll Pitch Yaw (RPY)
Names for rotation around an axis.
Roll: Side to side tilt
Pitch: Front to back tilt
Yaw. Direction pointing (think magnetic compass)
S
Satisficing
Satisficing allows for adequate rather than perfect results being acceptable. Joins satisfy and suffice.
Search Based Planning Library (SBPL)
A library with a motion planning approach using predefined motion primitives that are selected to be executed by the planning library.
Simultaneous Localization and Mapping (SLAM)
Simultaneously (so at the same time), building a map and localizing where in the map you are.
Singularity
Configuration where two joints become aligned in the same axis. This often causes difficulty controlling an axis since it might not move in the direction you want (or not move at all).
Sliding Mode Autonomy (SLMA, SMA)
The idea of having various levels of autonomy and the line between autonomy and human interaction changing.

SLMA should not be used as an abbreviation to avoid confusion with SLAM.
Super Odometry
Localization algorithm that is IMU centric and can use cameras or LIDAR to bound the errors.
Support Vector Machine (SVM)
Machine learning algorithm for regression or classification. Generally nice due to low computational cost.
T
Travelling Salesman Problem (TSP)
Quest to find the shortest (for whatever your cost function is) most efficient route for a salesmen/robot to travel given a list of goal locations. This is a common problem for robotics researchers (and others) to talk about and try to solve.
U
U-Net
A type of convolutional neural network for semantic segmentation
Unmanned Aerial Vehicle (UAV)
Drone, with no person onboard
Unmanned Ground Vehicle (UGV)
Ground robot, with no person onboard
V
Visual Inertial Odometry (VIO)
Method for determining position based on both visual (camera) sources and inertial (IMU) data.
Visual Odometry (VO)
Determining your position based on motion detected in visual sources (ie cameras).

Liked it? Take a second to support David Kohanbash on Patreon!
Become a patron at Patreon!

Leave a Reply