gego
An external attachable pad that acts as an interface between in illiterate worker and a sophisticated robot.
Core trick
Converts voice or text commands into robot motion through a safety layer that filters commands before they reach a ROS 2 robot.
Description
An external attachable pad that acts as an interface between in illiterate worker and a sophisticated robot. Accesible and comnandable by voice, text, the pad intelligently converts them to safe robot motion by also maintating iteractability.
README excerpt
<!-- mascot: a cartoon GECKO in Gemma/Gemini/Google colours (blue→purple→pink gradient
skin, cheeky grin, one hand waving). GeGo = "GemmaGO", and it sounds like "gecko". -->
<p align="center">
<img src="media/banner.jpg" alt="GEGO — GemmaGO: talk to your robot, it stays safe" width="100%">
</p>
<p align="center"><b>GemmaGO 🦎 — small brain. big impact. zero bumps.</b><br>
<i>(yes, it's pronounced like "gecko" — that's the mascot)</i></p>
---
## What is GEGO?
GEGO is a **plug-and-play brain-pack for robots** 🦎 — like a gecko, it clips onto *any* ROS 2
robot and, once attached, that robot inherits **voice control** and **formal safety** with no
per-robot tuning. A **non-expert can then drive it just by talking to it** — in English, Hindi,
or whatever they speak — and it stays safe while doing it. The robotics expertise is built in.
It runs a full **sense → think → act** loop **on-device and offline**:
- a **local multimodal LLM (Gemma)** is the *mind* — it sees, converses, and decides **what** to do,
- a **formal safety controller (Spatiotemporal Tubes)** is the *nervous system* — it decides **how**
to move and **guarantees** the robot avoids obstacles.
The guiding principle is **"the LLM never touches geometry."** A 4-billion-parameter model is great at
language and lousy at metric math — so everything metric (obstacle distances, goal positions, the
motion itself) is computed by deterministic code, and Gemma only does what it's good at: understand,
talk, and ch