Embodied Robot Interaction Framework — ChatPiano (MBZUAI Thesis)

Developing an embodied robot interaction framework based on multimodal large model agents, integrating unified vision-language-action representations. Implementing “one-sentence-multiple-step” long-horizon task closed-loop with rich dialogue experiences.

The system uses a non-blocking multi-threaded architecture with parallel tool servers and notification threads for real-time execution.

Performed alongside the China National Symphony Orchestra during their UAE arts festival visit at MBZUAI, demonstrating AI-human collaborative performance.