ETH Zurich and NVIDIA teach a quadruped robot to solve a Rubik's cube with its feet
lukas_m_ziegler · x · 2026-10-30
New work from ETH Zürich and NVIDIA Robotics shows a quadruped autonomously solving a giant Rubik's cube with its feet — a capability they call "pedipulation."
The robot walks around the cube and scans it with an onboard depth camera: SAM 3 segments the cube, ICP locks its 3D pose (2 cm position error, under 3° rotation), facelet colors are classified to reconstruct the full cube state, and a standard solver plans the moves, executed by learned skills.
The hardest skill is turning a layer: one foot pins the cube while the other pushes the top layers 90°, all while both legs keep the robot balanced. Skills are trained with RL in simulation, then distilled into policies running on just depth sensing plus the robot's joint encoders.
More from Embodied
- OpenFlyV2: a 1B onboard VLA model for real-time aerial drone navigation — IPEC-COMMUNITY · 2026-10-30
- Robots from Asimov and Menlo Research survive brutal hardware testing — alysha_lobo · 2026-10-30
- XPENG's IRON humanoid robot walks off its new production line, 80%+ core processes automated — ingliguori · 2026-10-30
- Lenovo ThinkSmart View boots mainline Linux 7.x, turning the $30 device into a Claude/OpenAI speaker — DavidBennett__ · 2026-10-30
- Amazon's DreamFAR world model nearly doubles π₀ real-world performance for robot training — CyberRobooo · 2026-10-30
- KAIST's L3G-World Generates Infinite Interactive Worlds in Real Time with Latent 3D Gaussian Memory — kaist-ai · 2026-10-30