Xiaomi introduced the four-legged robot companion CyberDog. Outwardly, it is similar to Boston Dynamics: instead of the head, there are several cameras and sensors, including the Intel RealSense D450 camera with depth detection, and other sensors are distributed on the body of the designed dog. These include touch sensors, ultrasonic sensors, and the Global Positioning System (GPS).
This is as it should be CyberDog can recognize and guide people, eg to help with pregnancy (maximum 3 kg). CyberDog can also do some tricks like backflips, otherwise it’s more than just a gimmick.
Data is processed by Nvidia’s Jetson Xavier NX AI platform in Xiaomi’s CyberDog, which uses six self-designed ARM cores (codenamed Carmel) and a Volta graphics unit with 384 shader cores and 48 tensor cores (total 21 TOPS KI computing power). TSMC, the chip-order manufacturer, is producing System-on-Chip (SoC) as of 2018 with 12nm technology. The Jetson Xavier NX can use two deep learning accelerators, but the computing power is not enough for complex real-time learning.
It is controlled by voice commands, remote control or smartphone app. CyberDog operates at speeds of up to 3.2 meters per second or 11.5 kilometers per hour. Xiaomi has not revealed how long it can keep up with the pace.
An open source operating system is preinstalled on a 128GB SSD and should be adaptable as needed. In the future, hardware modules that can be connected via three USB Type-C ports and an HDMI port, eg lidar for better navigation, should follow.
Initially, Xiaomi only sells 1,000 copies of CyberDog in China for the equivalent of 1,320 euros. Meanwhile, the manufacturer does not show photos or videos of the real robot – all offers have been made so far.
Similar robots are significantly more expensive, but they are also intended for industrial use, and not for private use. The basic version of Boston Dynamics Instant Program costs $74,500, plus thousands to tens of thousands of dollars for additional modules such as the Advanced Computing System for Local Accounts.