Skip to main content

G1 Architecture: Resolved Technology Decisions

Duration: 60 min · Level: Foundational · Module: 10. G1 Development Roadmap · Focus: G1, architecture, systems, decisions

Learning objectives

By the end of this lesson you will be able to explain and apply:

  • Actuation
  • Locomotion
  • Perception
  • AI brain
  • Hands

Why this matters​

Pulling together all modules, this lesson makes the final architecture calls for Autosapien G1.

Overview​

Pulling together all modules, this lesson makes the final architecture calls for Autosapien G1. Every decision is grounded in the state-of-the-art reviewed across this curriculum, with explicit rationale and known risks.

Key concepts​

Key idea

Actuation: QDD motors (locomotion joints) + SEA (arm joints) hybrid; HEBI Robotics X5-4 series for prototype; custom in-house actuators for production

  • Locomotion: RL-based policy trained in Isaac Lab + MPC fallback controller; target 1.5 m/s walking, 0.5 m/s stair climbing, 0.3 m/s backward walking
  • Perception: 4× RGB-D cameras (forward stereo + bilateral) + 2× fisheye cameras + 2× 6-axis IMU + 2× 6-DOF wrist F/T sensors + fingertip DIGIT tactile sensors
  • AI brain: dual AGX Orin setup (perception + VLA on one; locomotion + safety monitor on second); fine-tuned π0 or OpenVLA as manipulation policy backbone
  • Hands: custom 22-DOF tendon-driven design with DIGIT tactile on all 10 fingertips; total hand weight <800g per hand; peak fingertip force 20N
  • Power: 4.5 kWh LFP battery; 48V architecture; 8+ hours at moderate activity; hot-swap design in <5 minutes

Check your understanding​

Try to recall each answer before expanding it.

Q1. What do you know about Actuation?

QDD motors (locomotion joints) + SEA (arm joints) hybrid; HEBI Robotics X5-4 series for prototype; custom in-house actuators for production

Q2. What do you know about Locomotion?

RL-based policy trained in Isaac Lab + MPC fallback controller; target 1.5 m/s walking, 0.5 m/s stair climbing, 0.3 m/s backward walking

Q3. What do you know about Perception?

4× RGB-D cameras (forward stereo + bilateral) + 2× fisheye cameras + 2× 6-axis IMU + 2× 6-DOF wrist F/T sensors + fingertip DIGIT tactile sensors

Q4. What do you know about AI brain?

dual AGX Orin setup (perception + VLA on one; locomotion + safety monitor on second); fine-tuned π0 or OpenVLA as manipulation policy backbone

Q5. What do you know about Hands?

custom 22-DOF tendon-driven design with DIGIT tactile on all 10 fingertips; total hand weight <800g per hand; peak fingertip force 20N

Next: 10.2 Critical Risk Areas & Mitigation Plans →

Part of Module 10: G1 Development Roadmap.