Unitree Launches Next-Generation Go2 Quadrupedal Robot Platforms, Starting at $1,600

Robotics specialist Unitree has announced three new entries in its quadruped family, the Unitree Go2 Air, Pro, and Edu — the latter optionally including an on-board NVIDIA Jetson Orin NX computer-on-module (COM) with up to 100 tera-operations per second (TOPS) of sparse INT8 computer performance.

Building, as the name implies, on the existing Unitree Go platform, the Go2 quadrupedal robot offers a claimed 30 percent boost to motor performance over its predecessor, support for heat-pipe cooling to reduce operating temperature, a new vector positioning and control system including an "Intelligent Side-follow System 2.0," and a maximum running speed of up to around 11 miles per hour — in lab conditions, admittedly.

On the software side, Unitree's new robot, brought to our attention by Linux Gizmos, also offers support for Generative Pre-trained Transformer (GPT) large language model technology. "[This] enables Go2 to fully understand its user's intention," the company claims, "and better comprehend its surroundings."

Every model in the range offers a 720p front-facing camera with a 120 degree field of vision (FoV), plus a wide-angle 360×90° four-dimensional lidar sensor, and dual-band Wi-Fi 6 and Bluetooth 5.2 connectivity, with optional 4G cellular on the Pro and Edu models.

Where the models differ is in the details. The entry-level Air offers no built-in computing system, a 15.4-pound payload capacity, and a 5.6 mile per hour top speed, while lacking the vector positioning tracking system and auto-scaling payload strap of its higher-end alternatives.

The Pro, by contrast, adds the missing tracking system and strap, increases the payload capacity to 17.6lbs, boosts the top speed to around 7.8 miles per hour, and offers an unspecified "eight-core high-performance CPU" on board.

It's the top-end Edu that packs the most functionality, though. In addition to a boosted top speed of 8.3 miles per hour rising to 11 miles per hour in lab conditions, the Go2 Edu includes foot-end force sensors, an optional NVIDIA Jetson Orin Nano or NX system-on-module with 40 to 100 TOPS of sparse INT8 compute performance, and a 15Ah battery which boosts the run time from 1-2 hours to 2-4 hours.

More information on the new models is available on the Unitree website — but while the company has priced the Go2 Air and Go2 Pro at $1,600 and $2,800 respectively, it has left pricing for the top-end Go2 Edu model strictly on-application.