Bridging Robot-Free and Real-World Perception
Physical robots perceive the world through multiple viewpoints, typically a head-mounted camera for global context and wrist-mounted cameras for fine-grained manipulation. XRZero-G0 addresses this gap with a multi-view aligned sensing system.
Making Robot-Free Demonstrations Truly Trainable
XRZero-G0 formalizes trainability governance via a closed-loop Collection-Inspection-Training-Evaluation pipeline. Experiments show an effective data yield of around 85% under controlled experimental settings.
A 10:1 Mixing Law Reduces Real-Robot Requirements
Controlled experiments show that combining approximately 10 robot-free episodes with 1 real-robot episode achieves performance comparable to purely real-robot datasets. This strategy reduces the need for real-robot data by up to 20x.
G0-Dataset: 2,000+ Hours of Validated Data
Built on XRZero-G0, G0-Dataset provides over 2,000 hours of validated multimodal demonstrations spanning vision, tactile, and audio modalities.
Source: AIJourn/PRNewswire
