Nvidia researchers demonstrate significant progress in robot training
Researchers at Nvidia GEAR Lab and Carnegie Mellon University have developed a framework, ASAP (Aligning Simulation and Real Physics), that reduces the gap between simulated and real-world robot movements by 53%. The system trains robots in simulation and then uses a specialized model to adjust for real-world variations. In tests with the Unitree G1 humanoid robot, ASAP improved movement accuracy, enabling complex actions like jumps … Continue reading Nvidia researchers demonstrate significant progress in robot training
