Kyzen Builds on Stanford's Doggo, Pupper Platforms to Deliver a Low-Cost Robot with On-Device AI

Startup Kyzen Systems is looking to bring quadrupedal robot platforms to the masses — starting with the eponymous Kyzen, a low-cost robot with computer vision and visual simultaneous localization and mapping (vSLAM) built-in.

"We love quadruped robots, but almost all of them are either too small to do much or far too expensive," claims Raymond Mann of the reason for his company's formation. "That's why we've been working for the past three years to bring down the cost of practical quadruped robots."

The result of those three years: Kyzen, a low-cost smart robot platform capable of carrying a 2.2lbs payload. Built with three cameras on its "face," the robot comes equipped with face, object, and color tracking capabilities, plus the ability to follow a specific person — driven, its creators explain, by the YoLo neural network running entirely on-device.

The robot also boasts support for mapping, using visual simultaneous localization and mapping (vSLAM) with a forward-facing range-finder to create an internal map, which can be used to instruct the machine to visit specific rooms, route around obstacles, or find new obstructions in its environment.

The Kyzen platform itself is based on the work of Nathan Kau, creator of Stanford University's Doggo and Pupper — but Kyzen believes it has improved upon the original design. "Part of our work has included removing the need for a remote control and making Kyzen completely autonomous," Mann explains, "along with redesigning the legs and overall internals to last longer and go faster. This meant giving it a real brain, vision system built from three cameras, and an accelerometer."

The company is now crowdfunding Kyzen on Kickstarter, priced at $412 for early bird backers as a kit or $447 assembled. Hardware is expected to be delivered starting in September 2022.