Dextrous Robots Train in Simulation

⊲ WIRED ⊳ Eka Robotics in Cambridge have developed a robot that demonstrates unusually fluid, tactile manipulation, from grasping keys to screwing in a light bulb. Cofounders Pulkit Agrawal and Tuomas Haarnoja describe a simulation-led training approach that lets machines learn dexterity through repeated virtual practice rather than human demonstrations. Their vision-force-action model integrates touch and physics, aiming to narrow the gap between simulation and reality and enable broader use in industry, food handling, and daily tasks.
⊲ Image – WIRED ⊳
www.simulation.news