🐕 Researchers at the University of Hong Kong and Oxford Robotics Institute have developed ConsJump, a hierarchical reinforcement learning system that let a 48-pound quadruped robot autonomously assess and leap through narrow gates comparable to its own body size, reaching speeds of 2.5 meters per second before landing.
🔗https://link.ie.social/6Vv2Ho
Interesting Engineering
Quadruped robot learns to jump through tight gaps without human help
Four-legged robot learns dog-like agility to leap through narrow spaces using AI-powered vision and control.
4September 1, 2026 1.1K 2
