Physical AI
AI systems that perceive, reason, and act through physical machines in the real world.
Open topicCommunity research index
Learn how intelligent systems perceive, reason, and act in the physical world.
A structured roadmap through robotics, world models, embodied intelligence, and vision-language-action systems, built for students, builders, and researchers.
- 23
- core topics
- 30
- paper explainers
- 27
- research profiles
Interactive roadmap
Build knowledge in a deliberate sequence
Select a stage to see its role and open the full lesson.
Direct answer
What should you learn first in Physical AI?
Start with linear algebra, probability, Python, machine learning, computer vision, robotics, and control. Then study robot learning, world models, and vision-language-action systems in that order.
Core curriculum
Explore the field by topic
Foundations
The mathematics, programming, machine learning, vision, and robotics concepts needed to study Physical AI.
Open topicEmbodied Intelligence
Intelligence that emerges through an agent's body, sensors, actions, and interaction with an environment.
Open topicWorld Models
Learned models that predict environment dynamics and possible future outcomes.
Open topicVLA Models
Models that map visual observations and language instructions to robot actions.
Open topicRobot Learning
Methods that allow robots to acquire behavior from demonstrations, rewards, interaction, or generated experience.
Open topicResearch archive
Landmark papers, explained before linked
World Models
A compact latent model can let an agent learn behavior inside its own predicted environment.
World ModelsReinforcement Learning
Read explanation Dreamer: Reinforcement Learning with Latent Imagination
Dreamer learns long-horizon behavior by propagating value gradients through imagined latent trajectories.
World ModelsModel-Based RL
Read explanation DreamerV2
DreamerV2 extends latent imagination with discrete representations and reaches human-level Atari performance.
World ModelsAtari
Read explanation DreamerV3: Mastering Diverse Domains through World Models
DreamerV3 uses robust normalization and objectives to learn across more than 150 tasks with one configuration.
World ModelsGeneral Agents
Read explanation RT-1: Robotics Transformer for Real-World Control at Scale
RT-1 trains one transformer policy on a large multi-task dataset of real robot demonstrations.
VLAManipulation
Read explanation RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
RT-2 co-trains vision-language models on web and robot data so semantic knowledge can influence actions.
VLATransfer Learning
Read explanation Common questions
Frequently asked questions
What is Physical AI?
Physical AI is artificial intelligence that perceives, reasons, and acts through robots or other physical systems under real-world constraints.
Where should a beginner start?
Start with programming, linear algebra, machine learning, computer vision, robotics, and the perception-action loop before reading advanced VLA or world-model papers.
What is a vision-language-action model?
A vision-language-action model connects images and natural-language instructions to actions that a robot can execute.