ROS stack for cognitive machines
A ROS stack that uses AI for logic, not just motion—now with live Twitch demos.
A new ROS 2 stack called bob-ros2 takes a different approach to robotics—instead of focusing on motion control, it aims to build cognitive machines. The project treats ROS as the nervous system for AI-driven logic, providing a collection of packages that handle everything from deterministic filtering to semantic AI tasks. Key components include an LLM node that wraps OpenAI-compatible chat APIs, a Qwen3 TTS module, and an audio/video streamer originally built for Twitch but usable locally with ffmpeg.
The stack is actively maintained and evolves nearly daily, with extensive documentation in all packages. It can be seen live on Twitch, where users can interact with the system, ask questions, request real-time coding to solve tasks, or even let the AI evolve—all with human moderation. Most sub-components have their own repositories and can be used independently. The project connects to a broader ecosystem of ROS-AI integrations, including Agent ROS Bridge and AgenticROS.
- Uses ROS as a nervous system for AI logic, not physical movement
- Includes an OpenAI-compatible LLM node and Qwen3 TTS for conversational and audio capabilities
- Live on Twitch with interactive features like real-time coding and evolution, moderated
Why It Matters
Enables robotics developers to build cognitive AI systems using familiar ROS infrastructure and components.