The Chengdu Humanoid Robot Innovation Center (hereinafter referred to as the "Innovation Center") has released China's first multimodal robot model and dual arm collaboration system in Chengdu, Sichuan. This model enables robots to understand and reason abstract semantic instructions, and schedule dual arm collaborative systems to perform tasks. Multimodal data collaborative reasoning refers to the process of integrating two or more different perception sources, conducting comprehensive reasoning, and completing complex tasks. Researchers at the Innovation Center will apply multimodality to the field of humanoid robots, integrating various factors such as images, semantics, force perception, and environmental perception to comprehensively judge, generate tasks, and execute them. This is a key core technology for humanoid robots to have autonomous thinking ability. Zhang Ruirui, Chairman of the Innovation Center, introduced that semantic instructions include distinguishing and sorting beverages and non beverages relatively clearly, as well as distinguishing and sorting cylindrical and rectangular objects, edible and inedible items relatively abstractly. To complete the above tasks, the robot needs to autonomously extract and summarize all items through a multimodal model. Especially for multiple objects with similar bottle shapes, the robot needs to further determine whether they are edible items rather than non edible bottled items such as cosmetics based on the text, images, and other elements on the bottle packaging in order to complete the task. The release of multimodal models not only accelerates the process of 'AI physicalization', but also enables humanoid robots to evolve from 'preset posing' and 'remote control operation' to having stronger 'brains' and autonomous execution capabilities Zhang Ruirui stated that in the next step, the innovation center will continue to deepen technology research and development, continuously optimize the integration of multimodal models and dual arm collaboration systems, and enhance the perception, decision-making, and execution capabilities of humanoid robots in complex environments. (New Society)