VMIVME-5510 There are many areas where AI technology can be enabled. Why did you choose the subdivision of “Using AI technology to help 3D visually-enabled robotic arms complete flexible operation perception”?

J: The field of AI covers a wide range of fields, including computer vision, natural language processing, speech recognition, etc. I have been deeply engaged in the subfield of computer vision since 2000, and have accumulated considerable research achievements in the years of research. 3D vision, because of its advantages of imitating the human eye and forming stereovision in physical space, has become the lowest demand in computer vision. Since 2015, I have mainly focused on the field of 3D vision, integrating AI technology into the application of 3D vision. This study looked at the entire international academic circle at that time and was a relatively advanced existence.

In the many branch applications of AI technology to help 3D vision, after comprehensive consideration, enabling the robot arm to complete the relevant flexible operation perception is landable and the commercial realization cycle is short. And most of the scenes of industrial intelligent manufacturing are semi-open environments, VMIVME-5510 and through AI technology to help 3D vision, it can fundamentally change the inherent defects of the flexible operation of the previous robot arm in the traditional scheme. In the long run, such a product form can change the design ideas of the current automated production line, achieve the automation of the entire industry line by building blocks, and also push automation to a new stage of intelligent manufacturing.

M: In the limited time since the establishment of cross-dimensional intelligence, it has burst out unlimited potential. Can you share your company’s core technology?

J: Starting from the bottom, the core technology of cross-dimensional intelligence is Sim2Real AI cloud platform DexVerse. Most of the AI technology in the industry is still in the traditional sense of image recognition, understanding, semantic segmentation and other image levels. The 3D technology itself solves problems in three-dimensional physical space, perceives and locates the spatial position of objects, and completes relevant operations. The cross-dimensional DexVerse platform is a new type of AI platform. It can establish the mapping association between Sim2Real virtual space and real space by opening up VMIVME-5510 the closed loop of physical simulation, data synthesis, AI design and training, and model hardware and software deployment, and realize AI empowerment for tasks such as flexible operation of robot arm, robot perception and control.

For example, how a robotic arm picks up a glass of water on a table may seem like a breeze in human perception, but in the eyes of a machine, it’s a rather complicated process. The first thing you need to read is how far is the cup placed on the table from the robot arm? What’s the height? Secondly, the world’s cups are ever-changing, and after the shape of the cup changes, can the machine still recognize that it is a cup? How does the machine deal with “cups” of different shapes, colors, and sizes?

With the help of cross-dimensional cloud platform DexVerse, the robot arm can train various data needed to pick up the water cup through physical simulation and data synthesis to support the robot arm’s subsequent grasping, positioning, loading and unloading, unstacking and other operations of the target object. Simulate interactions, movements, and deformations between objects in a simulation space to predict physical behavior in the real world, and synthesize virtual data for deep learning model training.