// Workers AI · dad joke modeWhy did physical AI go to gym? To pump up its bytes.
Physical artificial intelligence or physical AI refers to artificial intelligence systems that perceive, reason about and act within the physical world. These systems generally combine AI models with sensors, control systems, actuators and physical machines such as robots or autonomous vehicles. Physical AI overlaps with embodied artificial intelligence, robotics and autonomous systems, but emphasizes the complete process of perceiving an environment, motion planning an action and physically executing the task to perform work, compared to digital AI or generative AI that primarily stays in the information or digital realm.[1][2][3]
The term became increasingly prominent during the AI boom in the 2020s as AI development expanded from primarily digital applications toward humanoid robots, self-driving vehicles, smart factories and other autonomous machines.[4] Its boundaries are not standardized, and it is often treated as a continuation of earlier research in robotics and embodied intelligence rather than an entirely separate field.[5]
Operation
[edit]
Physical AI systems commonly operate through a continuous cycle of perception, planning and action. Cameras, lidar, radar, microphones and tactile or motion sensors collect information about the environment. Computer vision, machine vision, sensor fusion and simultaneous localization and mapping may be used to identify objects, estimate their positions and construct a representation of the surrounding world.[6][7]
Open-source libraries such as OpenCV and Dlib provide computer-vision, image-processing and machine learning components that can be incorporated into perceptual systems. These libraries provide individual software components rather than complete autonomous systems.[8][9]
After interpreting its surroundings, a system may use task planning, motion planning or learned policies to select an action. Control software converts the plan into commands for robot locomotion for motors, robotic joints, or other actuators. New sensor information allows the system to evaluate the result and modify its plan as environmental conditions change.[10]
Applications
[edit]
Applications of physical AI include self-driving vehicles, industrial and warehouse robots, humanoid and service robots, drones, delivery robots, autonomous agricultural, construction and mining robots. Household applications include robotic vacuum cleaners and robotic lawn mowers that use sensors and navigation software to operate with limited human control.[11]
Physical AI systems must function under changing and partly unpredictable real-world conditions. Challenges include incomplete sensor data, collision avoidance, real-time computing requirements, energy constraints and transferring behavior learned in simulation to physical machines. Because failures can cause physical damage or injury, these systems may also require safety constraints, human oversight and fallback control mechanisms.[12][13]
See also
[edit]| Part of a series on |
| Artificial intelligence (AI) |
|---|
| Glossary |
- Autonomous robot
- Embodied agent
- List of artificial intelligence algorithms
- Lists of open-source artificial intelligence software
- list of open-source robotics hardware
- List of robotics companies
- List of robotics simulators
- List of robotics software
- Open-source robotics
- Robot learning
- Vision-language-action model
- World model (artificial intelligence)
References
[edit]- ↑ "What Is Physical AI?". NVIDIA. Nvidia. Retrieved 26 July 2026.
- ↑ Stryker, Cole (19 January 2026). "What Is Physical AI?". IBM Think. IBM. Retrieved 26 July 2026.
- ↑ "Black Forest Labs Unveils First Model for Robotics in Shift to Physical AI". Bloomberg News. 23 July 2026. Retrieved 26 July 2026.
- ↑ Cherney, Max A. (9 January 2026). "Physical AI dominates CES but humanity will still have to wait a while for humanoid servants". Reuters. Retrieved 24 July 2026.
- ↑ "From embodied intelligence to physical AI". Nature Machine Intelligence. 8: 491–492. 24 April 2026. doi:10.1038/s42256-026-01239-3.
- ↑ Dellaert, Frank; Hutchinson, Seth (2023). "Introduction". Introduction to Robotics and Perception. Retrieved 26 July 2026.
- ↑ Nahavandi, Saeid; Alizadehsani, Roohallah; Nahavandi, Darius; Mohamed, Shady; Mohajer, Navid; Rokonuzzaman, Mohammad; Hossain, Ibrahim (2025). "A Comprehensive Review on Autonomous Navigation". ACM Computing Surveys. 57 (9). doi:10.1145/3727642.
- ↑ Pulli, Kari; Baksheev, Anatoly; Kornyakov, Kirill; Eruhimov, Victor (June 2012). "Real-Time Computer Vision with OpenCV". Communications of the ACM. 55 (6): 61–69. doi:10.1145/2184319.2184337.
- ↑ King, Davis E. (2009). "Dlib-ml: A Machine Learning Toolkit". Journal of Machine Learning Research. 10 (60): 1755–1758. Retrieved 24 July 2026.
- ↑ Mon-Williams, Ruaridh; Li, Gen; Long, Ran; Du, Wenqian; Lucas, Christopher G. (19 March 2025). "Embodied large language models enable robots to complete complex tasks in unpredictable environments". Nature Machine Intelligence. 7: 592–601. doi:10.1038/s42256-025-01005-x.
- ↑ "What is Physical AI?". HPE. Hewlett Packard Enterprise. Retrieved 26 July 2026.
- ↑ Cherney, Max A. (8 July 2026). "Mistral launches first robotics model in physical AI push". Reuters. Retrieved 24 July 2026.
- ↑ "NVIDIA Announces Halos for Robotics, the Industry's First Full-Stack Safety System for Physical AI". NVIDIA Newsroom. Nvidia. 22 June 2026. Retrieved 26 July 2026.
