SwarmChat
Natural Language Control for Robot Swarms
Enabling intuitive human-swarm robot control through Large Language models and behavior trees.
Command robot swarms using Natural Language. No programming knowledge required - just speak or type your instructions naturally.
Integrated LLAMA Guard and behavior tree validation ensure safe, reliable execution. Malicious commands are automatically filtered out.
Supports multiple languages through advanced speech-to-text and translation models, making robotics accessible globally.
SwarmChat combines multilingual speech recognition, advanced language models, safety filtering, and structured behavior tree generation to create a robust human-robot interface.
These results reflect the performance of our fine-tuned model under zero-shot conditions. Fine-tuning with LoRA achieved dramatic improvements across all metrics, especially in zero-shot scenarios where no examples are provided.
Get started with SwarmChat by cloning our open-source repository and following the setup instructions.
GitHub RepoExplore our full research paper detailing SwarmChat’s architecture, methods, and performance results.
Read PaperExperience SwarmChat in action with our interactive demo showcasing natural language robot control.
Live Demo