Glossary

This glossary defines key terms used throughout the textbook "Physical AI & Humanoid Robotics".

Actuator: A component of a robot that converts energy into mechanical motion, enabling the robot to move or interact with its environment. Examples include electric motors, hydraulic cylinders, and pneumatic pistons.
AI Alignment: The field of research dedicated to ensuring that autonomous artificial intelligence systems act in accordance with human intentions, values, and ethical principles.
Artificial Intelligence (AI): The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
ASIMO (Advanced Step in Innovative Mobility): A humanoid robot developed by Honda, known for its advanced bipedal locomotion and human-like movements.
Autonomy: The ability of a robot or AI system to make decisions and operate independently without continuous human intervention.
Behavior-Based Robotics: An approach to robotics that emphasizes simple, reactive behaviors that emerge from direct interaction with the environment, often without complex internal representations or planning.
Bipedal Locomotion: The act of walking or moving on two legs, a complex challenge for humanoid robots.
Cobot (Collaborative Robot): A robot designed to work directly and safely with humans in a shared workspace, often without physical barriers.
Compliance: The ability of a robot to deform or yield in response to external forces, often achieved through passive mechanical design or active control strategies.
Degrees of Freedom (DOF): The number of independent parameters that are required to completely specify the configuration or motion of a robot or its components.
Dynamics: The branch of mechanics concerned with the study of forces and torques and their effect on the motion of objects, including robots.
Embodiment: The concept in AI and robotics that intelligence is not purely an abstract cognitive process but is deeply influenced and shaped by the physical body and its interaction with the environment.
End-Effector: The device or tool located at the end of a robotic arm, designed to interact with the environment (e.g., grippers, hands, welding torches).
Exteroceptive Sensors: Sensors that provide information about the robot's external environment, such as cameras, LiDAR, and tactile sensors.
Foundation Model: A large-scale AI model, typically pre-trained on vast and diverse datasets, that can be adapted (e.g., fine-tuned) for a wide range of downstream tasks. Examples include Large Language Models (LLMs) and Large Visual Models (LVMs).
Forward Kinematics: The calculation of the position and orientation of a robot's end-effector given the angles or positions of its joints.
Frame Problem: A philosophical and technical problem in AI concerning how to efficiently represent and update a robot's knowledge base about what changes and what stays the same in the world when an action is performed.
Grasp Planning: The process of determining how a robot should grip an object to achieve a stable and task-relevant grasp.
Human-Robot Interaction (HRI): The study of how humans and robots can interact with each other in a safe, efficient, and natural manner.
In-Hand Manipulation: The ability of a robot to reorient or move an object within its gripper without needing to release and re-grasp it.
Inverse Kinematics: The calculation of the required joint angles or positions for a robot to place its end-effector at a desired position and orientation.
Kinematics: The branch of mechanics concerned with the study of the motion of objects without reference to the forces that cause the motion.
Latent Representation: A compressed, lower-dimensional summary of high-dimensional data (e.g., an image) that captures its most salient features, often learned by AI models.
LiDAR (Light Detection and Ranging): A sensing technology that uses pulsed laser light to measure distances and create detailed 3D maps of the environment.
Linear Inverted Pendulum Model (LIPM): A simplified dynamic model used in bipedal locomotion, where the robot's mass is concentrated at a single point (center of mass) and the legs are massless, allowing for real-time walking pattern generation.
Model Predictive Control (MPC): An advanced control strategy that uses a model of the system to predict its future behavior and optimizes control inputs over a finite time horizon to achieve a desired goal while satisfying constraints.
Multi-modal Reasoning: The ability of an AI system to process, integrate, and reason about information from multiple sensory modalities, such as vision, language, and touch.
Perception-Action Loop: A continuous cycle in which an agent senses its environment (perception), processes the information (cognition/control), and acts upon the environment (action), with the actions influencing subsequent perceptions.
Physical AI: A subfield of AI focused on developing intelligent agents that are embodied, situated in the physical world, and learn through interaction.
Proprioceptive Sensors: Sensors that provide information about the robot's internal state, such as joint angles (encoders) or orientation (IMUs).
Reality Gap: The discrepancy between a robot's behavior in simulation and its behavior in the real world, often a challenge in sim-to-real transfer.
Reinforcement Learning (RL): A type of machine learning where an agent learns to make decisions by performing actions in an environment to maximize a cumulative reward.
Risk Assessment: A systematic process of identifying hazards, estimating risks, evaluating risks, and implementing risk reduction measures for a robotic system.
Robotics: The interdisciplinary branch of engineering and computer science that deals with the design, construction, operation, and use of robots.
Sensor: A device that detects and responds to some type of input from the physical environment (e.g., light, heat, motion, pressure) and converts it into a signal that can be read by a robot.
Shared Autonomy: A control paradigm where a human operator and an autonomous system collaborate on a task, each contributing to the control based on their respective strengths.
Situatedness: The concept that an intelligent agent's intelligence and behavior are deeply intertwined with its physical environment and context.
Symbolic AI: An early paradigm of AI that focuses on representing knowledge using symbols and manipulating those symbols with logical rules, often detached from physical interaction.
Tactile Sensor: A sensor that detects physical contact and measures properties such as pressure, force, and texture, providing a robot with a sense of touch.
World Model: An internal representation or simulation that an agent learns about its environment, enabling it to predict future states and imagine the consequences of its actions.
Zero Moment Point (ZMP): A concept used in bipedal locomotion to define the point on the ground where the net moment of all forces (inertial and gravitational) acting on the robot is zero, a key indicator of robot stability.