Kepler leverages NVIDIA physical AI to close industrial loop

Shanghai Kepler Robotics announced its industrial deployment pipeline based on NVIDIA's physical AI technology stack: using Isaac Lab for simulation training, GR00T and Cosmos for model training and deployment, then validating in real industrial scenarios such as workshop assembly and CNC machine loading/unloading, forming a closed loop from simulation to real machines. Kepler said the company achieves continuous Sim2Real iteration through a cycle of "simulation training → real-machine deployment → discrepancy analysis → parameter feedback → model correction → retraining," and introduces the real dynamic parameters of its self-developed planetary roller screws, actuators, joint modules and robot bodies into simulation. Kepler also said it has jointly developed an embodied intelligence foundation model for long-horizon tasks with the team from Zhejiang University's College of Artificial Intelligence. The above industrial scenario validation and model R&D progress all come from the company's own statements.

AI 根据报道生成 · 5 days ago更新