ETH Zürich Robot Swings Across Monkey Bars Using Raw Lidar, No Terrain Map
lukas_m_ziegler · x · 2026-09-12
The robot from ETH Zürich's legged robotics lab jumped onto monkey bars, brachiated across, and dropped down to a stable landing.
Why standard approaches fail: most legged robots build terrain maps first, but a heightfield discards exactly the geometry that matters here—thin overhanging bars with air above and below.
Technical approach:
- The policy reads raw scans from a head-mounted solid-state lidar, with no reconstruction step
- An attention-based encoder picks out which sparse returns matter
- A GRU carries memory through the sequence, counting when a target bar slides out of the sensor's view
Training: a phase-scheduled teacher-student setup with separate privileged experts for jumping up, brachiating, and jumping down—each with its own curriculum—and a scheduler handing control between them.
More from Embodied
- Regret optimal control: EPFL paper offers a metric for measuring robot stupidity — ducha_aiki · 2026-09-12
- GPT-6 Astra rumored to use recurrent-depth latent reasoning, sparking AI safety alarms — DJiafei · 2026-09-12
- Robot drags a 50kg rice bag needlessly — how do we measure robot "stupidity"? — ducha_aiki · 2026-09-12
- ECCV paper: one scalar per patch from pre-trained ViTs enables fast real-world robot navigation — chriswolfvision · 2026-09-12
- Magnotherm brings magnetic cooling fridges to REWE stores, using 15% less power — skdh · 2026-09-12
- Real2sim first, then real: user reproduces Astra controlling a robot arm to paint — ZeYanjie · 2026-09-12