Fly a Drone with a Phone Call

What if controlling a drone was as simple as making a phone call?

In this month's meetup, we’ll show how we built a voice-controlled autonomous drone system that lets you fly a UAV using natural language—no controller, no app, just your voice.
By combining Twilio for telephony, a large language model (LLM) for reasoning, and the Model Context Protocol (MCP) as a bridge to real-world actions, we translate spoken commands into safe, structured drone behaviors. Under the hood, a microservices architecture connects a voice pipeline to a ROS 2–powered autonomy stack, enabling real-time control and feedback.

You’ll see how:

• A phone call becomes a stream of AI-interpreted commands
• The LLM decides what to do (intent + reasoning)
• MCP translates that intent into executable “skills” like takeoff, movement, and scanning
• A ROS 2 autonomy agent safely flies the drone (or simulation)

We’ll walk through the architecture, key design decisions, and challenges of bridging AI with physical systems—latency, safety, and reliability included.