Brian Metrillo
Work

Multi-drone control interface

A tablet app for one operator to fly a small team of drones indoors by dragging them on a live map.

Role
Software co-lead
Year
2026
Stack
TypeScript, Python, ROS
Code
Source

Overview

This is my Computer Engineering thesis, and it is still in progress. The goal is a tablet app that lets one operator fly a small team of drones indoors by dragging each one on a live map. I co-lead the software side. I wrote the message format between the tablet and the flight computer, the Python command bridge, and the page for the first test flights.

Problem

The thesis asks whether one person can move several drones at once with touch on a map, plus hand gestures for height. That only works if the software stays safe when a finger lifts, the tablet drops off the network, or the radio goes quiet.

Approach

We split the system so most of it runs without drones. The tablet web app and the bridge talk through a written message format, and the bridge reaches drones only through one drone connection interface. A mock of that interface returns fake positions, so the tablet, the bridge, and their tests run on a laptop.

Each safety rule, like what a released drone or a stop does, went into a short decision record before the code.

How it works

The tablet draws each drone at its measured position and sends the operator's touches to the bridge. The bridge checks every message, keeps a target for each flying drone, and streams those targets whether or not anyone is touching the screen.

  1. 1The operator taps take off, and the drone climbs to a low, fixed height.
  2. 2A finger on the drone holds it, and dragging moves its target on the map.
  3. 3The tablet refuses any drag into the restricted zone before it sends the target.
  4. 4The bridge checks the target again and streams it to the drone connection.
  5. 5When the finger lifts, the drone holds the spot where it was released, and land brings it down.

If the tablet disconnects or input goes stale, every drone keeps its last target, and control stays locked until the operator lifts every finger. Stop drops every hold and keeps the drones where they are. The connection to real drones also limits how fast a target can move and lands any airborne drone when the bridge shuts down, though so far only against a fake.

Testing

One contract test suite runs against the mock drone connection and against the real one with a fake robotics client. The fake copies firmware behavior we read in its source, like cutting the motors when commands stop. Other tests open a real socket to the bridge and build their messages from shared example files. The bridge can also log every input and target for the test flights.

Outcome

Against the mock, the tablet and the bridge work end to end, from take off to one finger drags, stop, and land. The real drone connection is written but has not flown a drone yet. Next are the first single drone test flights, then several drones at once and gesture control for height.