Training autonomous robots for unpredictable real-world environments requires vast amounts of data, making purely physical testing costly and risky. Siemens‘ Simcenter Engineering Services, in collaboration with EPFL‘s Automatic Control Laboratory, aims to address this by combining digital twin simulation, built using Simcenter Amesim with real-world data. Robots are first exposed to varied simulated scenarios covering terrain, weather, and lighting. Notably, the approach requires only minimal real-world data to fine-tune robot behavior. The decision-making module continuously updates as new data is gathered, with AI tools like Simcenter Studio Reinforcement Learning expected to further advance autonomous robot development.
Click to read the original article..
Image generated by: ChatGPT
