Xiaomi has developed the Xiaomi-Robotics-1 model using over 100,000 hours of motion data collected from humans operating camera-equipped handheld grippers. According to The Decoder, this approach of leveraging human-generated motion data rather than robotic data has significantly enhanced the robot's motion capabilities.
The company found that increasing the volume of training data had a more substantial impact on improving robot motion performance than simply scaling up the model size. This insight highlights the importance of quality and quantity of data in advancing robotics technology.
For Japanese markets, where robotics and automation continue to be key growth sectors, Xiaomi’s data-driven approach underscores the competitive edge gained through extensive real-world data collection in AI and robotics development.