Zero Shot Sim to Real

Zero-shot sim-to-real transfer aims to bridge the gap between simulated and real-world robotic environments, enabling robots trained solely in simulation to perform tasks in the real world without further real-world training. Current research focuses on improving simulation realism through techniques like 3D Gaussian splatting and diffusion models, and developing robust policy architectures such as language-conditioned transformers and those leveraging pre- and post-contact decomposition, to handle the inherent differences between simulated and real-world physics and sensor data. Success in this area is crucial for accelerating robotics development, reducing the cost and risk associated with real-world training, and ultimately enabling wider deployment of robots in diverse and unpredictable environments.