Hong Kong – Recently, xiaomi (1810.HK) released its first open source quadruped robot CyebrDog, with 11 high-precision sensors and self-developed high-performance servo motors to sense images, light, distance, speed, sound and other environmental information, allowing it to follow instructions, automatically avoid obstacles, complete various high-speed movements and backflips, and other difficult movements. Highly difficult movements, and can even identify the master and automatically follow the master movement. At the same time, users can wake up CyberDog by voice to control more than 72 categories and 3,000 smart devices.
The robot is called “CyebrDog” in xiaomi’s internal project, and is currently limited to 1,000 units, priced at 9,999 yuan.
Trade Date: 2021-08-13, Xiaomi Corporation,(HKSE:1810.HK)
Open: $26.05, Close: $25.45, Change: $-0.60
Low: $25.20, High: $26.20
Year High: $35.90, Year Low: $12.88
Tages: 1810.HK, Xiaomi Corporation, Consumer Electronics, Technology
CyberDog configuration parameters (source: xiaomi)
CyebrDog’s core is Nvidia’s world’s smallest embedded and edge system artificial intelligence processor Nvidia JetsonXavier NX and Intel Real Sense D450 depth module, supplemented by six microphones, and then realize the robot’s perception of the world.
The Nvidia Jetson Xavier NX contains the performance of 384 CUDA cores, 48 Tensor cores, 6 Carmel CPU cores and two NVIDIA Deep Learning Accelerator (NVDLA) engines, and in the case of CyebrDog, this processor module is the central brain for the 11 high-precision sensors mentioned above. Measuring just 70 mm x 45 mm, the Nvidia Jetson Xavier NX delivers 21 TOPS of accelerated computation while running modern neural networks in parallel and processing data from multiple high-resolution sensors.
Nvidia Jetson Xavier NX (Source: Nvidia)
Intel Realtime D450, effective distance of 0.3-6 meters, global shutter, accuracy error of less than 2.2% within 7 meters. Can be 90 frames per second frequency output 1280 x 720 resolution of the depth of the stream. For the computer vision system to build a complete 3D coordinate system.
Intel Real D450 and vision processor D4 (source: Intel)
Generally speaking, it is mainly industrial cameras that can be equipped with Nvidia Jetson Xavier NX, such as the industrial camera NEON-2000-JNX series launched by ADLINK in May 2021. This is because the core advantage of this processor is a large reduction in integration complexity and provides great reliability verification operations, and for CyebrDog, the robot is in a scenario where it is difficult to perfectly play the advantages of the processor.
Similarly, the Intel Real D450 is also commonly used in industrial inspection and other scenarios. These two products, plus CyebrDog’s electric heart that can reach 220rpm, almost make the BOM price reached or even exceeded the $9,999 selling price. Is xiaomi really just trying to make an overpriced advertisement for the “technology open source community”?
Not really, the quadruped bionic robot in now has long been not surprising. But for robot integration, traditional robots require complex integration of image sensor modules, cables and GPU modules, on the other hand, because the camera can collect 3D coordinate vision, a large amount of raw data will lead to algorithms to reduce the efficiency of operation. Ultimately, integrators must change the traditional integration method to a smaller volume to complete the acquisition to the edge and then to the network of agile integration.
To achieve this integration technology, you need to understand the different electronic components, how to make these embedded components work together, and also learn how to develop and optimize algorithms for different platforms such as Nvidia and Intel, and so on. Take the hottest fully autonomous mobile robots (AMR) in the field of robotics today, such as Vecna Robotics, ForwardX Robotics, etc., the integrated technology power has become their core competency. These companies have some experience in developing different hardware platforms to ensure that after the launch of new hardware, the overlapping generation can be achieved quickly and new AMRs can be launched on the market.
Robotics is an applied science, and the logic of xiaomi’s CyebrDog is very clear: first take the strongest hardware platform, and then gradually complete the skilled application of the platform through integration. Once it has the industrial demand, xiaomi can rely on the integration experience gained from the development of CyebrDog for rapid development and implementation, and finally complete the layout of the high-end integration industry.
Industry Insights Dept
TickerInsider News Network