If you are interested in machine learning, then you are probably at least familiar with the Kendryte AI-accelerated processors. The Kendryte K210 is a RISC-V dual core 64-bit processor running at 400 MHz. Enhancements aimed at speeding up AI-related tasks allow the K210 to perform up to a trillion operations per second.
One of the more intriguing applications of this processor is a robotic dog called the XGO-Mini, developed by Luwu Intelligence Technology. This tiny quadruped robot’s brain is a Kendryte K210 processor. The head also contains a camera, as well as a graphical display to provide feedback to users. A nine-axis inertial measurement unit and joint positioning sensors give the robot information about its environment to aid in smooth, natural movement.
This little robot comes packed with features. It is capable of face detection, image recognition, object tracking, voice recognition, human posture recognition, dynamic gestures, and traffic sign recognition, for starters.
There is good news for those that do not want to spend hours assembling their new pet — XGO-Mini comes fully assembled. It is controlled with a smartphone app, but can also recognize and interact with objects and gestures that it detects with its camera. For those that want to hack some new functionality into the bot, it is possible to program it with Python.
A major upgrade from the K210 processor was recently announced — the Kendryte K510 tri-core RISC-V AI processor. Running at twice the clock speed of the K210, the K510 can perform up to three trillion operations per second via its AI processing subsystem. Adding in 1.5 megabytes of SRAM and a video processing unit, this is a powerful, general purpose, platform for edge AI applications. Perhaps a future version of the robot could take advantage of it.
The XGO-Mini recently went live on Kickstarter. If you are quick, it can still be had for $599, with a full retail price expected to be $1,098.