ETH Zurich teaches a robotic hand to walk on its own fingers with RL
CyberRobooo · x · 2026-09-16
ETH Zurich researchers taught the Wuji Hand to move autonomously by treating its fingers as legs: a reinforcement learning policy coordinates the fingers for locomotion, while onboard sensing and control let it adapt in the real world. The hand can crawl across gravel, lawns and stone paving, and even play golf and games. This points to robots moving and operating in narrow spaces—under desks, inside equipment gaps, even under beds—and possibly detached robotic hands executing tasks on their own.
Related event: ETH Zurich Teaches Robotic Hand to Walk on Its Fingers(5 posts)→
More from Fun
- Codex User Down to 3% Weekly Quota: 'The Real Race Condition' — EricBuess · 2026-09-16
- 15-second OS install demo, echoing dhh: 'Almost all constraints are imaginary' — Yamapama · 2026-09-16
- Dev builds website of 25 mini rooms where Claude keeps visitors company, made with Claude Opus 5 — repligate · 2026-09-16
- gnukeith clarifies viral clip is fully AI-generated fake: "I would never say this" — gnukeith · 2026-09-16
- 12-day-old AI agent emails professor seeking paid work to fund its own tokens — Polymarket · 2026-09-16
- Alignment researcher: bragging your AI might go out of control won't sell to enterprises — Turn_Trout · 2026-09-16