Glossary of Physical AI and Humanoid Robotics Terms
A
Admittance Control: A control strategy that regulates the relationship between applied force and resulting motion. The dual of impedance control, admittance control specifies how the system moves in response to external forces.
Adaptive Control: Control systems that adjust their parameters or structure in real-time based on changing system dynamics or operating conditions.
Affordance: The possibility of an action that the environment offers to an agent. In Physical AI, objects afford different types of interactions based on their physical properties.
Attractor: A set of states toward which a dynamical system evolves over time. Point attractors, limit cycles, and chaotic attractors represent different types of stable behaviors.
B
Bifurcation: A qualitative change in the behavior of a dynamical system that occurs when system parameters cross critical thresholds.
Bio-Inspired Robotics: Robotics approaches that draw inspiration from biological systems, including their morphology, control strategies, and learning mechanisms.
Body Schema: An internal representation of the robot's body configuration and its relationship to the environment, essential for coordinated movement and interaction.
Braitenberg Vehicle: A simple robot architecture demonstrating how basic sensor-motor connections can produce complex behaviors that appear to show emotions like fear or love.
C
Capture Point: The location where a robot must step to come to a complete stop, considering its current momentum and balance state.
Centroidal Dynamics: The dynamics of a robot's center of mass and centroidal momentum, used for balance and locomotion control.
Central Pattern Generator (CPG): A neural network that produces rhythmic patterns without rhythmic input, often implemented as coupled oscillators for robotic locomotion.
Compliance Control: Control strategies that regulate the mechanical impedance of a robot to achieve desired force-motion relationships during interaction.
Contact Dynamics: The study of forces and motions that occur when two or more bodies come into contact, including impact, friction, and constraint forces.
Control Barrier Function (CBF): A mathematical tool that ensures safety constraints are satisfied by constraining control inputs.
Control Lyapunov Function (CLF): A Lyapunov function that is used in control design to ensure stability of the closed-loop system.
D
Dynamical Systems: Mathematical models describing how system states evolve over time, fundamental to understanding emergent behaviors in Physical AI.
Dynamic Balance: Maintaining balance during motion, as opposed to static balance which applies to stationary systems.
E
Emergent Behavior: Complex behaviors that arise from the interaction of simple rules with the environment, rather than being explicitly programmed.
Embodied Cognition: The theory that cognitive processes are deeply influenced by the body's interactions with the physical environment.
Environmental Coupling: The tight relationship between a physical system and its environment that contributes to intelligent behavior.
Event-Based Control: Control strategies that trigger updates based on events rather than fixed time intervals.
F
Feedback Control: Control systems that use measurements of system state to adjust control actions and correct errors.
Feedforward Control: Control actions based on anticipated disturbances or desired trajectories rather than measured errors.
Force Control: Control strategies that regulate the forces applied by a robot during interaction tasks.
Forward Dynamics: Computing motion from applied forces, as opposed to inverse dynamics which computes forces from desired motion.
G
Gait: The pattern of leg movements during locomotion, including walking, running, and other forms of legged movement.
Gaussian Process: A probabilistic model used for uncertainty quantification and learning in robotics applications.
Generalized Coordinates: A set of parameters that uniquely define the configuration of a mechanical system.
H
Hierarchical Control: Control architectures with multiple levels of abstraction, typically including high-level planning and low-level execution.
Hybrid System: Dynamical systems with both continuous and discrete dynamics, common in robotic systems with contacts.
Hybrid Zero Dynamics: A framework for stabilizing periodic orbits in systems with hybrid dynamics, particularly useful for walking robots.
I
Impedance Control: A control strategy that regulates the mechanical impedance (resistance to motion) of a robot to achieve desired interaction behaviors.
Impulsive Dynamics: Mathematical models that describe the instantaneous change in velocities during impacts.
Inverse Dynamics: Computing forces from motion, as opposed to forward dynamics which computes motion from forces.
Inverted Pendulum: A pendulum with its center of mass above its pivot point, used as a fundamental model for balance control.
In-Hand Manipulation: Manipulation of objects using controlled motion of the hand without releasing the object.
L
Learning from Demonstration: Machine learning approaches that acquire behaviors by observing expert demonstrations.
Limit Cycle: A closed trajectory in phase space that represents a stable periodic behavior.
Linear Inverted Pendulum Mode (LIPM): A simplified model of balance dynamics with constant height, resulting in linear equations of motion.
Lyapunov Stability: A mathematical framework for analyzing the stability of dynamical systems.
M
Model Predictive Control (MPC): An advanced control strategy that solves an optimization problem at each time step to determine optimal future actions.
Morphological Computation: The contribution of physical body properties to intelligent behavior, reducing computational requirements on the controller.
Multi-Contact: Scenarios involving simultaneous contact at multiple points on a robot's body.
P
Passivity-Based Control: Control strategies that leverage the energy properties of physical systems to achieve stable behaviors.
Physical AI: The field of artificial intelligence focused on systems that interact with and operate in the physical world.
Port-Hamiltonian Systems: A mathematical framework for modeling physical systems that explicitly represents energy flow and interconnection structure.
Proportional-Integral-Derivative (PID) Control: A common feedback control algorithm with three terms that respond to error, error accumulation, and error rate of change.
R
Reinforcement Learning: A machine learning paradigm where agents learn to take actions in an environment to maximize cumulative reward.
Resonance: The phenomenon where a system oscillates with greater amplitude at certain frequencies, important in compliant robotic systems.
Robust Control: Control strategies designed to maintain performance despite uncertainties and disturbances.
S
Sensorimotor Coordination: The tight coupling between sensing and motor control that enables effective interaction with the environment.
Series Elastic Actuator (SEA): An actuator design that includes a spring in series with the motor and load, providing compliance and force sensing.
Sliding Mode Control: A variable structure control strategy that forces the system state to evolve on a predefined surface.
Stability: The property of a system to return to an equilibrium state after small disturbances.
State Estimation: The process of determining the internal state of a system from sensor measurements and system models.
Subsumption Architecture: A behavior-based control architecture that layers simple behaviors with higher-level behaviors inhibiting lower-level ones.
V
Variable Stiffness Actuator (VSA): An actuator that can actively adjust its mechanical impedance during operation.
Viscoelasticity: The property of materials that exhibit both viscous and elastic characteristics when deformed.
Virtual Fixtures: Virtual constraints or guides that assist human operators or autonomous systems in performing tasks.
Z
Zero Moment Point (ZMP): The point on the ground where the moment due to inertial and gravitational forces exactly cancels the moment due to ground reaction forces, used for balance analysis in walking robots.