← back to space

“Bro, Look at My Wheat” — HackHarvard

Text your drone to survey wheat, spot crop issues, and get field insights back on your phone — iMessage → Photon bridge → Tello drone, with NOAA/NASA weather cross-checks.

PythonDJI TelloPhotonElevenLabsYOLOArduino

Overview

Checking on crops should be as simple as sending a text. Instead of making a farmer learn a drone-control interface, the interaction starts in iMessage: text "Start survey," the drone takes off, flies its survey, lands, and the results come back to your phone — a crop-health insight, a JPEG, and optionally a voice note. Built at HackHarvard 2026 as an end-to-end hardware prototype: phone, command bridge, laptop, and a physical drone working as one system.

How It Works

  • iMessage commands ("start survey," "abort," "land") pass through an authenticated, opt-in Photon command bridge to the laptop ground station
  • A DJI Tello flies the survey; local YOLO wheat-head detection and AprilTag tracking handle steering
  • A persistent Factory Droid SDK session acts as the runtime mission brain: checks mission status, issues drone commands, and assesses survey images
  • Chalk joins crop scores with NOAA weather and NASA POWER satellite sunlight to justify each recommendation with environmental context, not just the photo
  • An Arduino UNO Q streams DHT11 temperature/humidity and LDR light readings as telemetry — sensors inform the insight but can never start a mission
  • Results return over Photon: the insight, a JPEG, and an optional ElevenLabs voice note for critical alerts

My Contributions

  • Implemented the Photon communication layer — the authenticated iMessage command bridge in both directions: inbound commands ("start survey," "abort," "land") and outbound results with the survey image
  • Implemented the ElevenLabs text-to-speech path, so critical alerts come back as a voice note instead of just text

Challenges

  • Making separate pieces behave like one system: a phone request, command processing, and a physical drone all have to agree on when a survey starts and how its results get back to the user
  • Testing with real flights: every software change had to stay aligned with live drone tests while teammates built different parts in parallel
  • Showing it clearly: the phone recording, laptop recording, and camera footage started at different times, so we synchronized all three to make the end-to-end interaction visible

Project Media

Demo: "Start survey" sent from the phone, the Tello flying its survey, and the insight + voice note arriving back in iMessage.

Links

Tools & Methods

PythonDJI TelloPhoton 2.0ElevenLabsFactory Droid SDKChalkYOLO → ONNXNOAA / NASA APIsArduinoSoldering

Built at HackHarvard 2026.

recruiter mode →