Wearable system captures human activity data for robot training
RoboticsMechanical Engineering
THE AI ANGLE
Estimating unmeasured body joints and training simulated humanoid controllersResearchers built a wearable sensor rig and recorded a 100-hour dataset of human task demonstrations called EgoExplorer. The accompanying pipeline estimates full-body poses from sparse tracking points, giving roboticists synchronized vision, motion, and touch data to train simulated humanoid controllers.
THE TEACHING ANGLE
Students can examine whether motion inferred from sparse body trackers can transfer successfully to physical robots without sacrificing hand stability and precise contact.Read the original at techxplore.com Generate teaching or study materials
More in Robotics
- Sterling Anderson doesn’t think humanoid robots should be building carsThe Verge · October 27, 2026
- ANYbotics begins delivering ANYmal X legged robot for explosive atmospheresThe Robot Report · October 27, 2026
- Artificial skin helps robots grasp pencil lead and pull away from needlesPhys.org — Technology · October 27, 2026
- AI-powered robotic laboratory paves the way for discovering next-generation heat-resistant coatingsPhys.org — Chemistry · October 27, 2026
- Xpanner launches autonomous pile driver for solar farm contructionThe Robot Report · October 27, 2026