Shending Technology: The first 3D spatial computing chip supplier

On May 17, 2024, the "14th Songshan Lake China IC Innovation Summit Forum" hosted by the Integrated Circuit Design Branch of the China Semiconductor Industry Association and VeriSilicon Microelectronics (Shanghai) Co., Ltd. was held in Songshan Lake, Dongguan. Yuan Diwen, Chairman and CEO of Shending Technology (Nanjing) Co., Ltd., introduced the 3D spatial computing chip VC6801 for embodied intelligence at the forum.

Shending Technology was founded in 2020 and the company is registered in Nanjing. It has teams mainly in Shanghai, Nanjing, and Shenzhen. The team size is nearly 100 people, and R&D accounts for more than 80%. CEO Yuan Diwen is a veteran in the communications industry and has worked at MediaTek and Unisoc for more than 20 years. He said that Shending Technology is the first mass-produced supplier of embodied intelligent 3D spatial computing chips. The embodied intelligent terminal forms include industrial robots, autonomous driving, humanoid robots, low-altitude flight, AGV/AMR, and MR/AR.

"If we want embodied smart terminal devices to understand the world, the core technology is spatial computing, which requires better three-dimensional perception and sensor fusion capabilities," said Yuan Diwen.

The VC6801 3D spatial computing chip of Shending Technology can realize rich functions including 3D reconstruction, depth estimation, spatial positioning, perception fusion, etc. It supports a variety of positioning technologies including RTK, VIO, WiFi, UWB, as well as 3D perception technologies such as binocular, millimeter wave, ToF, and lidar. It has the advantages of strong adaptability, support for low-cost sensors, and low-power processing for heterogeneous sensor fusion. It has been successfully applied to various types of robots, naked-eye 3D notebooks, 3D fitting mirrors, AR helmets and other devices.

Yuan Diwen took the robot vacuum cleaner as an example to introduce the advantages of embodied intelligence in detail. For example, in terms of cleaning function, 2D image recognition cannot detect certain low-feature-value dirt, while 3D perception can increase the multi-dimensional information of objects and achieve more accurate cleaning. In terms of navigation/obstacle avoidance, laser radar requires additional modules, so the passability is poor and the product appearance is limited. However, Shending Technology's AI VSLAM can optimize the machine appearance, improve the pass rate, and save BOM costs.

Using mobile phones as an analogy, Yuan Diwen emphasized that the development of embodied intelligence also requires collaboration across the industry chain, and that chip companies also need to provide turnkey solutions including SDKs and solutions to accelerate the maturity of the robotics industry.