Hiranya Agarwal

Project 02 of 06 / Autonomy

ADA²M autonomous drone

A two-person project developing an agricultural monitoring drone. I work on the mechanical package and autonomy integration. The aircraft is flying, with stability tuning and sensor integration in progress.

Role
Mechanical and autonomy
Team
Two people
Period
2025 to now
Status
Flight testing
The assembled ADA2M quadcopter on grass during outdoor testing, showing its frame, battery and propellers.
The assembled airframeOutdoor testing of the current aircraft.

Flight test

From the workbench to short flights.

The flight-test footage shows the aircraft lifting from the grass and flying close to the ground. It documents an early flight test, not a completed autonomous monitoring mission.

Outdoor flight test, 25 secondsSilent footage. Playback starts only when you choose to play.
Read the video description

The quadcopter lifts a short distance above a grass lawn. Several views show the aircraft moving close to the ground and resting on the grass between attempts. The video demonstrates take-off and low-level flight, without measured endurance or stability results.

Hiranya working on the drone electronics with a soldering iron at a workbench.
Electronics integration on the benchWorking on the assembled airframe before outdoor testing.

The brief

Collect useful environmental data without walking the whole property.

Small scale farmers cannot instrument every corner of a paddock, and spot readings from one weather station do not describe a field. The goal is for ADA²M to follow a predefined coverage route and log environmental measurements along it. The engineering problem is packaging: an airframe, a flight controller, a long range link, onboard compute and a sensor payload all have to fit inside a mass and power budget that does not move.

Architecture

The intended monitoring and control stack.

The two control loops on the aircraft, and which one flies today Two nested loops. The inner loop is stabilisation and it is the part that flies today: the SpeedyBee F405 flight controller reads its inertial measurement unit, drives the electronic speed controllers and motors, and reads the result again many times a second. A TBS Crossfire link carries pilot commands and telemetry over long range. The outer loop is autonomy and it is still being integrated: a Jetson Nano is intended to hold the coverage route and run inference on board, handing the flight controller where to go rather than how to stay level, and a sensor payload is intended to log temperature, humidity and air quality along the route. Those two parts are drawn with dashed outlines because they are not proven yet. Ground stationpilot, telemetry F405 controllerstabilisation ESCs + motors Airframecarbon fibre IMUattitude and rate Jetson Nanoroute, inference Sensor payloadtemperature, humidity, air TBS Crossfire where to go inner loop: how to stay level motion dashed: designed, not proven in flight yet red: the loop that is flying today

Scroll sideways to see the whole diagram

Two loops, different jobsThe flight controller only has to keep the aircraft level and pointed, and it does that from its own inertial measurements. That loop is what flies today. The Jetson Nano never touches it: its job is deciding where the aircraft should be next, which is the slower problem and the one still being integrated.
Airframe
Custom carbon fibre frame
Long range link
TBS Crossfire
Design tool
SolidWorks
Software
C++ and Python

Engineering work

Mechanical packaging and staged integration.

  1. Package the flight stack. Design the carbon fibre frame around the propulsion, control and compute hardware rather than fitting the hardware to a stock frame.
  2. Connect autonomy to flight. Integrate Jetson Nano inference with onboard decision making so the aircraft can act on what it sees.
  3. Design for range. Use a TBS Crossfire link for long-range control and telemetry. The operating envelope still requires flight validation.
  4. Test in steps. Move from short hops to stability tuning before expanding sensor and autonomy capability, so changes can be evaluated incrementally.

Programme

Developed with Project Umbrella Labs.

ADA²M runs under Project Umbrella Labs, the student engineering startup I co-founded at UNSW. The lab works across autonomous UAVs, neural interfaces and agricultural monitoring, and runs its projects as rapid prototype, test, revise cycles.

projectumbrellalabs.com

Status

Airborne, and still being tuned.

Current status

The aircraft flies and is in stability tuning, with sensor integration continuing alongside it. Endurance, payload and coverage figures will be published once the test programme produces numbers I can stand behind.