Preparation curriculum · Simulation only

Learn robotic manipulation
from the ground up.

Short, practical lessons connected to a real MuJoCo project. Start with the feedback loop, then learn how frames make robot motion precise.

Current scope. This public site teaches the project’s preparation material. It does not claim a completed dinner-table robot, a trained policy, or Intel validation.

Learning path

  1. Lesson 01 — Observe, act, and step time
    Understand observations, actions, control intervals, physics steps, and what a runtime check establishes.
  2. Lesson 02 — Frames and reachable positions
    Use an interactive two-link sketch to see how coordinates, frames, and reachability relate.
  3. Lesson 03 — Contacts, grasps, and hand-offs
    See why a grasp and hand-off require physical evidence, not a convincing animation.
  4. Lesson 04 — Demonstrations and ACT
    Learn the aligned observation-action records that a bounded learned policy needs.
  5. Lesson 05 — Supervision and recovery
    Separate visual reasoning, bounded action policies, and the component that can stop.
  6. Lesson 06 — Evaluation and uncertainty
    Freeze test conditions and report every outcome before trusting a score.
  7. Lesson 07 — OpenVINO and benchmarks
    Measure actual device selection and complete application timing on Intel hardware.

Keep nearby

Open the product demo · Robotics vocabulary reference · Training evidence reference

Build with the project

The lessons correspond to a local, native-Apple-Silicon MuJoCo learning lab. The exact setup, evidence boundaries, and current implementation status live in the source repository.