This site is easiest to use when you start from an immediate goal instead of scanning the whole catalog in order.
The groups below — Learn first, Build systems, Prepare for real-world work, Track the field, and Use the project ladder — are a docs-site overlay on top of the repository's 14 canonical categories. The README itself does not contain "Learn", "Build", or "Deploy" headings; the cards link to the canonical category page on this site, with the equivalent section in the root README listed underneath.
Learn first
For: students, new joiners, engineers from adjacent fields, and managers ramping up.
If you are new to Physical AI or Embodied AI, start with structured material and move from there into papers.
CoursesStructured starting points include Stanford CS 336, Stanford CS 224R, Berkeley CS 287, CMU 16-831, and MIT 6.4210. README: #courses.BooksProbabilistic Robotics, Modern Robotics, Planning Algorithms, and related texts when you need slower, foundational coverage.Tutorials & GuidesLeRobot, Isaac Lab, MuJoCo, ROS 2, Open X-Embodiment, and Diffusion Policy links provide hands-on routes.Key PapersHighlights including V-JEPA 2, Open X-Embodiment, DROID, Diffusion Policy, RT-2, Octo, and other landmark entries.
Build systems
For: ML engineers, robotics researchers, and platform leads choosing tools, models, or infrastructure.
If you are choosing tools, models, or infrastructure, work through the canonical Build-side categories.
Robotics Foundation ModelsCompare entries like π0, V-JEPA 2, Octo, OpenVLA, RT-2, RT-X, Gemini Robotics, and NVIDIA GR00T. README: #robotics-foundation-models.World ModelsUse this branch when planning, prediction, or synthetic environment generation matters more than direct policy execution. README: #world-models.SimulatorsPhysics engines, high-fidelity simulators, and RL-focused environments. README: #simulators.ManipulationMethods, models, and tools for grasping, dexterous manipulation, and contact-rich tasks. README: #manipulation.LocomotionLegged, bipedal, and humanoid locomotion — controllers, learning approaches, and reference platforms. README: #locomotion.DatasetsOpen X-Embodiment, DROID, RoboMIND, and related training datasets. README: #datasets.
Prepare for real-world work
For: production-leaning engineers, integration teams, and the safety/governance owners attached to a deployment.
When you are moving from research tooling toward physical systems, route through the canonical Deploy-side categories.
Sim-to-RealDomain randomisation, system identification, and case studies for transferring policies to real hardware. README: #sim-to-real.Safety & RobustnessSafe-control references, guardrails, and robustness testing. README: #safety--robustness.Evaluation MethodologyHow to design and run trustworthy evaluations for physical systems. README: #evaluation-methodology.BenchmarksCALVIN, LIBERO, RLBench, HumanoidBench, and other comparable harnesses. README: #benchmarks.Production PatternsReference architectures, operational patterns, and deployment-adjacent references. README: #production-patterns--reference-architectures.Governance & PolicyStandards, regulation, and policy-adjacent material that shapes deployment. README: #governance--policy.Hardware PlatformsArms, humanoids, mobile robots, and lower-cost or DIY hardware separated for procurement and experiment planning.
Track the field
For: technical leaders, analysts, investors, and contributors who need a reading rhythm rather than a one-time onboarding.
If your goal is ongoing awareness instead of immediate implementation, work from the appendices and the Companies category.
Conference mapCoRL, RSS, ICRA, IROS, NeurIPS, ICML, ICLR, HRI, and Humanoids cover the major venues listed in the README.Community channelsROS Discourse, Robotics Stack Exchange, subreddit communities, and Discord servers provide discussion entry points.CompaniesActive commercial and research-product organisations in Physical AI. README: #companies.Newsletters & BlogsUse these when you want analysis, technical commentary, and field updates without reading every primary source.
Use the project ladder
For: anyone who learns by building — students, hobbyists, and engineers who want hands-on milestones rather than reading lists.
The Getting Started Projects section is the repository's most direct path from reading to practice.
BeginnerCart-Pole, browser MuJoCo, and introductory deep RL material keep the first steps lightweight.IntermediateLeRobot, Diffusion Policy, MuJoCo humanoid control, Isaac Lab, and RoboMimic push toward simulation and imitation work.AdvancedOpenVLA fine-tuning, Octo training, Open X-Embodiment, and RT-X evaluation move into current foundation-model workflows.Optional hardwareSO-100 Arm, ALOHA, Reachy Mini, and JetBot provide concrete physical build targets once simulation is no longer enough.