Bruncsoft-H1 (Hybrid 1)

A wheeled + legged humanoid robot, delivered as a fully articulated MuJoCo simulation.

Author: Viktor Brunclík · Licensed CC BY-NC-ND 4.0 — attribution required, non-commercial, no derivatives. See LICENSE.

Bruncsoft-H1 is a hybrid mobile base: it drives on a four-wheel platform on flat ground, and deploys retractable legs to climb stairs, rising from 1.20 m (wheels only) to 1.60 m (legs extended). It carries a 27-DOF right hand with a human-hand joint layout, a black carrying tray on the left forearm, six cameras, and sensors throughout.

---

Anatomy

Region	Description
Head	Rounded, 1X-style. Two eye-level cameras (eye_left, eye_right) behind a dark visor. Neck: yaw + pitch.
Torso	Fixed to the wheeled platform; shoulder yoke carries both arms. IMU in the chest.
Right hand	27 DOF, laid out like a human hand (breakdown below). Touch sensors on every fingertip + palm.
Left arm	Simple end effector plus a black plastic carrying tray mounted behind the elbow — 20 × 15 cm, with 5 cm walls.
Platform	Drives on four wheels (independently driven). Houses the legs, sensors and charger.
Legs	Four retractable telescoping legs deploy downward from the platform for stairs. Front legs carry the two forward cameras (cam_fl, cam_fr); back legs carry the two rearward cameras (cam_bl, cam_br).
Charger	Apple-Watch-style magnetic puck docked on the front-left leg.
Height	1.20 m wheels-only → 1.60 m legs extended.

The 27 degrees of freedom of the right hand

Matching the human hand:

Group	DOF	Joints
Wrist	3	flexion/extension, radial/ulnar deviation, pronation/supination
Thumb	5	CMC abduction, CMC flexion, MCP flexion, MCP abduction, IP flexion
Index	4	MCP abduction, MCP flexion, PIP, DIP
Middle	4	MCP abduction, MCP flexion, PIP, DIP
Ring	4	MCP abduction, MCP flexion, PIP, DIP
Little	4	MCP abduction, MCP flexion, PIP, DIP
Ulnar palm	3	transverse palmar arch + ring & little carpometacarpal (CMC) cupping
Total	27

The middle metacarpal is the fixed axis of the hand, so only the ring and little carpometacarpal joints (the mobile ulnar side) plus the palmar arch are modelled — exactly as in human hand kinematics.

Sensors

• IMU (accelerometer + gyro + orientation) on the base and the head
• Touch sensors on every fingertip, the palm, the carrying tray and all four feet
• Joint-position sensors on the finger MCPs
• Six cameras (2 head, 2 front-leg, 2 rear-leg)

---

Run it

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# scripted demo: drive → deploy legs → grasp with the hand → relax
python view.py

# with a staircase to climb
python view.py --scene

# or just open the interactive viewer on the raw model
python -m mujoco.viewer --mjcf model/bruncsoft_h1.xml

On macOS the MuJoCo viewer needs mjpython:

mjpython view.py

Files

model/
  build_model.py     parametric MJCF builder (edit this, then regenerate)
  bruncsoft_h1.xml   generated model — 45 actuators, 6 cameras, 19 sensors
  scene.xml          the robot + a staircase
view.py              interactive viewer / scripted demo

The model is generated:

python model/build_model.py   # rewrites model/bruncsoft_h1.xml

---

License

© 2026 Viktor Brunclík. Released under CC BY-NC-ND 4.0.

You may view, run and share this simulation with attribution to Viktor Brunclík. You may not use it commercially, and you may not distribute modified versions. See LICENSE for the full terms.