Learning Path · Advanced · Hardware Required

Booster K1: From Power-On to Embodied AI Experiments

A structured, safety-first path that takes you from your first K1 power-on to running whole-body imitation learning experiments on a full-size humanoid. Prerequisites: ROS2 basics, Python, and familiarity with robot kinematics.

Total Time ~20 hours
Difficulty Advanced
Hardware Booster K1 + Ethernet cable + safety mat
Prerequisites ROS2 basics, Python 3.10, robot kinematics familiarity
People Required Minimum 2 (operator + spotter)
You Will Build A whole-body manipulation demo with a trained policy

Before You Start — Prerequisites and Safety

  • Comfortable with ROS2 Humble — topics, nodes, launch files
  • Python 3.10+ and command-line proficiency
  • Familiarity with robot kinematics (joint space vs task space)
  • Ubuntu 22.04 installed natively on a host PC
  • A minimum 3 m × 3 m clear indoor space with rubber or foam flooring
  • A second person available as a spotter for all Unit 2+ sessions
  • Emergency stop hardware tested and within reach

Start with Unit 0 — it is mandatory safety orientation. You will not be allowed to power on the K1 until every checklist item is confirmed.

Your Path at a Glance

Complete the units in order. Each unit ends with a completion check. Do not proceed until you pass it.

0

Safety & Orientation Required

Emergency stop setup, workspace clearance, what NOT to do, and why this robot is different from a desktop arm. No SDK content yet.

~45 min · No code
1

Boot & Basic Motion Required

Power on, network configuration, SSH in, SDK install, DAMP mode, and first joint state read. Spotter required from this unit onward.

~2.5 hours
2

Locomotion Control Required

DAMP → PREP → WALK sequence, velocity commands, lateral movement, rotation, safe shutdown. First real walking session.

~3 hours
3

Upper Body Teleoperation Required

VR or leader-follower arm control setup. Head pose control, arm gestures, and CUSTOM mode (with lifting fixture). First teleoperation session.

~4 hours
4

Collect Whole-Body Demonstrations Required

Multi-camera setup, k1_agent.py recording, dataset format (30+ DoF), quality filtering, and safe session management.

~4 hours
5

Train & Deploy Required

Whole-body imitation learning with Diffusion Policy or ACT. Training on your recorded dataset, evaluation metrics, and live deployment with safety monitoring.

~6 hours

Time Breakdown

Unit Topic Time
Unit 0Safety & Orientation~45 min
Unit 1Boot & Basic Motion~2.5 h
Unit 2Locomotion Control~3 h
Unit 3Upper Body Teleoperation~4 h
Unit 4Whole-Body Data Collection~4 h
Unit 5Train & Deploy~6 h
Total ~20 h 15 min
Start Unit 0 — Safety & Orientation →

Unit 0 takes about 45 minutes. It is mandatory before you power on the K1 for the first time.