For our projects, we are always looking for a stable and capable platform to mount, test and develop new hardware.
In practice, this has not been easy. Many robotic platforms are either too fragile, too small, or too weak to operate in the kind of real-world environments that human operators are used to working in.
The Unitree Go2 Pro gives us a compact but serious platform to experiment with. It is mobile, relatively strong, and advanced enough to serve as a base for further development.
To be clear: the specific Go2 Pro model we currently have access to does not have an official IP rating. That means it should not be treated as waterproof, dustproof, or suitable for harsh outdoor conditions without additional protection.
Our first goal is to connect directly to the robot through its API. This allows us to read as many internal metrics as possible, including system status, movement data, sensor information and other available telemetry.
The second goal is to interpret those metrics using large language models. Instead of only displaying raw data, we want to explore how AI can help explain the robot’s condition, behaviour and possible issues in a more human-readable way.
The third goal is more ambitious: operating the robot with AI assistance. This means using AI not only to understand the robot, but eventually to help control it, guide it and make decisions based on live feedback from the platform.
This project is still in an early stage, but the direction is clear: we want to turn the Unitree Go2 Pro into a practical development platform for robotics, AI and field-operation experiments
