rai

Why we included this project

Teams building robots that should understand and act on natural-language commands will find RAI a practical starting point. It is an agentic framework that sits on top of your existing ROS 2 stack, so you can add LLM-driven reasoning, tool use, and multimodal perception to a Franka arm, a ROSbot, or a custom platform without rewriting your control layer. The framework handles the messy parts of embodied agents — turning a spoken or typed request into a sequence of ROS 2 actions, summarizing logs, and coordinating multiple agents — and it ships with demos for manipulation, navigation, and agricultural scenarios that show what a working setup looks like. Because it is vendor agnostic and modular, you can adopt just the pieces you need and keep the rest of your robotics infrastructure intact. It is a good fit for robotics engineers and researchers who want to prototype embodied AI quickly while staying close to the ROS ecosystem.