Physical AI & Humanoid Robotics Glossary
Version: 1.0 Total Terms: 24 Citation Style: APA
Purpose
This glossary provides canonical, unambiguous definitions of core domain terms for:
- Physical AI Handbook (academic textbook)
- Research paper scoping
- RAG-powered chatbot system
A
Actuation
Domain: Robotics
Definition: The mechanism by which a robot converts control signals into physical motion or force. Actuation systems include motors, hydraulics, pneumatics, or artificial muscles that enable robots to interact mechanically with their environment.
Source: Siciliano, B., Sciavicco, L., Villani, L., & Oriolo, G. (2009). Robotics: Modelling, Planning and Control (pp. 99-145). Springer. https://doi.org/10.1007/978-1-84628-642-1
Notes: Actuator selection impacts robot performance: electric motors offer precision, hydraulics provide high force, and pneumatics enable compliance. Humanoid robots commonly use electric actuators.
Related Terms: Actuator, End Effector, Motor Control
B
Bipedal Locomotion
Domain: Robotics
Definition: The mode of movement in which a robot walks on two legs. Bipedal locomotion in humanoid robots involves dynamic balance, gait generation, and coordination of multiple degrees of freedom to achieve stable, efficient walking on various terrains.
Source: Vukobratović, M., Borovac, B., Surla, D., & Stokić, D. (2012). Biped Locomotion: Dynamics, Stability, Control and Application (pp. 1-50). Springer. https://doi.org/10.1007/978-3-662-04740-9
Notes: Bipedal walking is inherently unstable (unlike quadrupedal locomotion) and requires active balance control. Key concepts include Zero Moment Point (ZMP), Center of Pressure (CoP), and gait patterns.
Related Terms: Humanoid Robot, Gait Generation, Zero Moment Point
C
Control Policy
Domain: Control
Definition: A mapping from states (or observations) to actions that defines the behavior of an autonomous agent. In robotics, a control policy specifies what actions the robot should take given its current sensory input and internal state. Policies can be deterministic or stochastic, handcrafted or learned.
Source: Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning: An Introduction (2nd ed., pp. 3-21). MIT Press. ISBN: 9780262039246
Notes: In classical control, policies are often called "controllers" or "control laws." In reinforcement learning, policies are typically denoted π and may be represented as neural networks, lookup tables, or parametric functions.
Related Terms: Controller, Feedback Control, Policy Gradient
D
Dynamics
Domain: Robotics
Definition: The study of forces and torques that cause motion in robotic systems. Robot dynamics describes the relationship between applied forces/torques, joint accelerations, velocities, and positions, accounting for inertia, friction, gravity, and external forces.
Source: Siciliano, B., Sciavicco, L., Villani, L., & Oriolo, G. (2009). Robotics: Modelling, Planning and Control (pp. 147-218). Springer. https://doi.org/10.1007/978-1-84628-642-1
Notes: Robot dynamics can be derived using Lagrangian or Newton-Euler formulations. Essential for model-based control, simulation, and motion planning with dynamic constraints.
Related Terms: Forward Dynamics, Inverse Dynamics, Newton-Euler Formulation
E
Embodied Intelligence
Domain: AI
Definition: The theory and practice of intelligence that arises from the dynamic interaction between an agent's body, brain, and environment. Embodied intelligence posits that intelligent behavior emerges from the coupling of perception, action, and environmental dynamics, rather than from abstract symbol manipulation alone.
Source: Pfeifer, R., & Bongard, J. (2006). How the Body Shapes the Way We Think: A New View of Intelligence (pp. 1-15). MIT Press. https://doi.org/10.7551/mitpress/3585.001.0001
Notes: Rooted in cognitive science and robotics. Challenges traditional cognitivist views of intelligence. Central to understanding Physical AI and humanoid robotics.
Related Terms: Physical AI, Morphological Computation, Situatedness
F
Feedback Control
Domain: Control
Definition: A control strategy that uses sensor measurements of system output to adjust control inputs, thereby reducing the error between desired and actual system behavior. Feedback control enables robustness to disturbances and model uncertainty.
Source: Åström, K. J., & Murray, R. M. (2008). Feedback Systems: An Introduction for Scientists and Engineers (pp. 1-24). Princeton University Press. ISBN: 9780691135762
Notes: Fundamental to robotics. Common implementations include PID control, state feedback (LQR), and model predictive control (MPC). Contrasts with feedforward control, which does not use sensor feedback.
Related Terms: Control Policy, PID Control, Closed-Loop Control
Forward Kinematics
Domain: Robotics
Definition: The problem of determining the position and orientation of a robot's end-effector given the joint angles or positions. Forward kinematics maps from joint space to task space using geometric and trigonometric relationships defined by the robot's mechanical structure.
Source: Craig, J. J. (2017). Introduction to Robotics: Mechanics and Control (4th ed., pp. 61-98). Pearson. ISBN: 9780133489798
Notes: Forward kinematics has a unique solution for a given joint configuration. Typically computed using Denavit-Hartenberg parameters or geometric methods.
Related Terms: Inverse Kinematics, End Effector, Joint Space
H
Humanoid Robot
Domain: Robotics
Definition: A robot with a body shape resembling the human form, typically including a torso, head, two arms, and two legs. Humanoid robots are designed to operate in human-centric environments and interact naturally with humans and human-designed tools and spaces.
Source: Siciliano, B., & Khatib, O. (Eds.). (2016). Springer Handbook of Robotics (2nd ed., pp. 1307-1333). Springer. https://doi.org/10.1007/978-3-319-32552-1
Notes: Humanoid form factor enables operation in human environments without environmental modification. Degrees of anthropomorphism vary from basic bipedal structure to highly detailed human-like appearance.
Related Terms: Bipedal Robot, Android, Anthropomorphic Robot
I
Imitation Learning
Domain: AI
Definition: A machine learning approach where an agent learns to perform tasks by observing and mimicking expert demonstrations, rather than through trial and error. In robotics, imitation learning enables faster learning from human demonstrations or teleoperation.
Source: Osa, T., Pajarinen, J., Neumann, G., Bagnell, J. A., Abbeel, P., & Peters, J. (2018). An algorithmic perspective on imitation learning. Foundations and Trends in Robotics, 7(1-2), 1-179. https://doi.org/10.1561/2300000053
Notes: Also called "learning from demonstration" (LfD) or "programming by demonstration." Common methods include behavioral cloning, inverse reinforcement learning, and generative adversarial imitation learning.
Related Terms: Behavioral Cloning, Learning from Demonstration, Teleoperation
Inverse Kinematics
Domain: Robotics
Definition: The problem of determining the joint angles or positions required to achieve a desired end-effector pose. Inverse kinematics is essential for task-space control, enabling robots to reach specified positions and orientations in the workspace.
Source: Craig, J. J. (2017). Introduction to Robotics: Mechanics and Control (4th ed., pp. 99-140). Pearson. ISBN: 9780133489798
Notes: Inverse kinematics may have multiple solutions, no solution, or infinite solutions (redundant manipulators). Common solution methods include analytical, numerical (Jacobian-based), and optimization approaches.
Related Terms: Forward Kinematics, Redundancy Resolution, Jacobian
M
Middleware
Domain: Robotics
Definition: Software that provides common services and capabilities to applications beyond those available from the operating system. In robotics, middleware facilitates communication between distributed components, manages data flow, and provides abstraction layers for hardware and software modules.
Source: Mohamed, N., Al-Jaroodi, J., & Jawhar, I. (2008). Middleware for robotics: A survey. 2008 IEEE Conference on Robotics, Automation and Mechatronics, 736-742. https://doi.org/10.1109/RAMECH.2008.4681463
Notes: Robotics middleware examples include ROS, ROS 2, YARP, and OROCOS. Middleware enables modularity, reusability, and distributed system development.
Related Terms: ROS 2, Communication Framework, Service-Oriented Architecture
Motion Planning
Domain: Robotics
Definition: The computational problem of finding a collision-free path or trajectory for a robot to move from an initial configuration to a goal configuration while satisfying kinematic, dynamic, and environmental constraints.
Source: LaValle, S. M. (2006). Planning Algorithms (pp. 1-38). Cambridge University Press. https://doi.org/10.1017/CBO9780511546877
Notes: Classical approaches include sampling-based methods (RRT, PRM), optimization-based methods (trajectory optimization), and graph-search methods. Modern approaches incorporate learning and real-time adaptation.
Related Terms: Path Planning, Trajectory Optimization, Collision Avoidance
P
Perception
Domain: Perception
Definition: The process by which a robotic system acquires, processes, and interprets sensory information from its environment to build internal representations for decision-making and control. In robotics, perception encompasses sensor data acquisition, signal processing, feature extraction, and scene understanding.
Source: Thrun, S., Burgard, W., & Fox, D. (2005). Probabilistic Robotics (pp. 93-152). MIT Press. ISBN: 9780262201629
Notes: Distinct from human perception in emphasizing computational processes and engineering constraints. Critical for autonomous operation in unstructured environments.
Related Terms: Sensor Fusion, Computer Vision, Exteroception
Physical AI
Domain: AI
Definition: Artificial intelligence systems that interact with and learn from the physical world through embodied agents, combining perception, reasoning, and actuation to perform tasks in real-world environments. Physical AI encompasses the integration of machine learning, computer vision, control theory, and robotics to enable autonomous physical systems.
Source: Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed., pp. 968-1012). Pearson. https://doi.org/10.1017/9781108608008
Notes: Emerging term that distinguishes embodied AI systems from purely virtual AI. Sometimes called "Embodied AI" in literature. Physical AI emphasizes the physical instantiation and real-world interaction capabilities.
Related Terms: Embodied Intelligence, Embodied AI, Robotics
R
RAG (Retrieval-Augmented Generation)
Domain: AI
Definition: A hybrid AI architecture that enhances large language models by retrieving relevant information from external knowledge bases before generating responses. RAG systems combine dense retrieval (finding relevant documents) with generative models to produce factually grounded, contextually relevant outputs.
Source: Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459-9474. arXiv:2005.11401
Notes: RAG addresses hallucination problems in LLMs by grounding generation in retrieved facts. Essential for building reliable AI assistants and chatbots in specialized domains like robotics and physical AI.
Related Terms: Language Model, Vector Database, Semantic Search
Reinforcement Learning (Physical Systems)
Domain: AI
Definition: A machine learning paradigm where an agent learns to make sequential decisions by interacting with a physical environment, receiving rewards or penalties based on its actions. In physical systems, RL must address challenges including sample efficiency, safety, partial observability, and continuous high-dimensional state-action spaces.
Source: Kober, J., Bagnell, J. A., & Peters, J. (2013). Reinforcement learning in robotics: A survey. The International Journal of Robotics Research, 32(11), 1238-1274. https://doi.org/10.1177/0278364913495721
Notes: Physical RL differs from simulated RL due to safety constraints, expensive real-world samples, and stochasticity. Common approaches include policy gradient methods, actor-critic algorithms, and model-based RL.
Related Terms: Policy Gradient, Sim-to-Real Transfer, Robot Learning
ROS 2
Domain: Robotics
Definition: Robot Operating System 2: An open-source middleware framework for robot software development, providing communication infrastructure, device drivers, libraries, and tools for building complex robot systems. ROS 2 is a complete redesign of ROS 1, offering real-time capabilities, improved security, and better support for multi-robot systems.
Source: Macenski, S., Foote, T., Gerkey, B., Lalancette, C., & Woodall, W. (2022). Robot Operating System 2: Design, architecture, and uses in the wild. Science Robotics, 7(66), eabm6074. https://doi.org/10.1126/scirobotics.abm6074
Notes: Industry standard for robot software development. ROS 2 uses DDS (Data Distribution Service) for communication, enabling deterministic real-time performance. Widely adopted in research and industry.
Related Terms: Middleware, DDS, ROS 1
S
Sensor Fusion
Domain: Perception
Definition: The process of combining sensory data from multiple sensors to produce more accurate, reliable, and comprehensive information than could be obtained from any individual sensor. Sensor fusion algorithms handle uncertainty, redundancy, and complementary information across different sensor modalities.
Source: Durrant-Whyte, H., & Henderson, T. C. (2016). Multisensor data fusion. In B. Siciliano & O. Khatib (Eds.), Springer Handbook of Robotics (2nd ed., pp. 867-896). Springer. https://doi.org/10.1007/978-3-319-32552-1_35
Notes: Common techniques include Kalman filtering, particle filters, and Bayesian inference. Critical for robust perception in robotics.
Related Terms: Perception, Kalman Filter, Multimodal Sensing
Sim-to-Real Gap
Domain: AI
Definition: The discrepancy between simulated and real-world performance when transferring models, policies, or control strategies learned in simulation to physical robots. The gap arises from imperfect modeling of physics, sensors, actuators, and environmental factors.
Source: Zhao, W., Queralta, J. P., & Westerlund, T. (2020). Sim-to-real transfer in deep reinforcement learning for robotics: A survey. 2020 IEEE Symposium Series on Computational Intelligence (SSCI), 737-744. https://doi.org/10.1109/SSCI47803.2020.9308468
Notes: Mitigation strategies include domain randomization, system identification, reality gap modeling, and hybrid sim-real training. Critical challenge for scaling robot learning.
Related Terms: Reinforcement Learning (Physical Systems), Domain Randomization, Transfer Learning
SLAM
Domain: Perception
Definition: Simultaneous Localization and Mapping: The computational problem of constructing or updating a map of an unknown environment while simultaneously tracking the agent's location within that map. SLAM is fundamental to autonomous navigation in unknown environments.
Source: Cadena, C., Carlone, L., Carrillo, H., Latif, Y., Scaramuzza, D., Neira, J., Reid, I., & Leonard, J. J. (2016). Past, present, and future of simultaneous localization and mapping: Toward the robust-perception age. IEEE Transactions on Robotics, 32(6), 1309-1332. https://doi.org/10.1109/TRO.2016.2624754
Notes: SLAM is a chicken-and-egg problem: localization requires a map, and mapping requires knowing position. Modern solutions use probabilistic frameworks (EKF-SLAM, FastSLAM, Graph-SLAM) or learning-based approaches.
Related Terms: Localization, Mapping, Visual SLAM
T
Trajectory Optimization
Domain: Control
Definition: A method for finding optimal robot motions by formulating motion planning as a mathematical optimization problem, minimizing a cost function (e.g., time, energy, smoothness) subject to dynamic constraints, kinematic limits, and environmental obstacles.
Source: Kelly, M. (2017). An introduction to trajectory optimization: How to do your own direct collocation. SIAM Review, 59(4), 849-904. https://doi.org/10.1137/16M1062569
Notes: Common methods include direct collocation, differential dynamic programming (DDP), and sequential quadratic programming (SQP). Enables dynamic, efficient, and natural robot motions.
Related Terms: Motion Planning, Optimal Control, Direct Collocation
V
Vector Database
Domain: AI
Definition: A specialized database optimized for storing and querying high-dimensional vector embeddings, enabling efficient semantic similarity search. Vector databases are essential for RAG systems, recommendation engines, and similarity-based retrieval tasks.
Source: Johnson, J., Douze, M., & Jégou, H. (2019). Billion-scale similarity search with GPUs. IEEE Transactions on Big Data, 7(3), 535-547. https://doi.org/10.1109/TBDATA.2019.2921572
Notes: Common vector databases include FAISS, Pinecone, Qdrant, Weaviate, and Milvus. They use approximate nearest neighbor (ANN) algorithms like HNSW or IVF for efficient retrieval.
Related Terms: RAG (Retrieval-Augmented Generation), Embedding, Semantic Search
Z
Zero Moment Point (ZMP)
Domain: Robotics
Definition: A point on the ground where the net moment of all forces acting on a bipedal robot is zero in the horizontal plane. ZMP is a key stability criterion for bipedal walking: if the ZMP lies within the support polygon, the robot will not tip over.
Source: Vukobratović, M., & Borovac, B. (2004). Zero-moment point—Thirty five years of its life. International Journal of Humanoid Robotics, 1(1), 157-173. https://doi.org/10.1142/S0219843604000083
Notes: ZMP criterion is widely used for planning and controlling bipedal gaits. Alternative stability criteria include Foot Rotation Indicator (FRI) and Capture Point.
Related Terms: Bipedal Locomotion, Center of Pressure, Support Polygon
Glossary Statistics
- Total Terms: 24
- Domains:
- AI: 6 terms
- Robotics: 12 terms
- Control: 4 terms
- Perception: 3 terms
- Citation Coverage: 100%
- Cross-References: All terms linked
Last Updated: 2024-12-24 Version: 1.0