Project 02

Custom Autonomous Quadcopter

A quadcopter designed from nothing: frame, wiring, and flight software. Five CAD iterations from a frame that snapped in five minutes to one that holds a stable hover and tracks a target on its own.

Fusion 360
Raspberry Pi
OpenCV
pymavlink

5frame iterations

100%custom CAD frame

Pi + camonboard compute

ELRSradio link

The brief

Design and CAD-model a fully custom autonomous quadcopter frame, reinforce it to survive crashes, integrate a flight-controller / ESC stack, brushless motors, and an ELRS radio system, and hand-solder every power and signal connection.

Then diagnose the flight-stability problems that showed up (frame flex, weight balance, oscillation) and fix them through redesign until the airframe hovers cleanly with a Raspberry Pi, camera, and GPS aboard.

Stack

Design	Fusion 360, 3D-printed frame
Flight controller	MicoAir F405 V2 flight stack (FC + ESC)
Motors	Brushless, 1000 KV
Radio	ExpressLRS (ELRS)
Companion	Raspberry Pi + USB camera + GPS
Autopilot	ArduPilot (MAVLink / GUIDED)
Vision	Python, OpenCV

The iterations

v1 → v5

Iteration 1

Baseline airframe

Quadcopter frame, iteration 1

First printable frame: top plate, bottom plate, and four arms. The goal was just to prove the geometry, motor mounts, and stack layout worked together.

Held the motors and the full flight stack solidly, and made thrust on a ground tether, a strong start
Arms and body were too thin: the frame snapped within five minutes of testing
No landing legs, so it sat flat on the floor and never got a clean lift-off

Frame size	500 mm motor-to-motor
Outcome	Did not fly; frame far too flexible

Iteration 2

Thicker frame + landing legs

Quadcopter frame, iteration 2

Thickened the arms and the body, and added leg stilts so the battery and the Raspberry Pi could mount underneath the drone.

Thicker arms and body
New: landing legs, so the battery and Pi now mount below the frame
Still snapped: thick enough in CAD, but the print infill and wall pattern were too weak to back it up

Changes vs v1	Arm thickness, body thickness, landing legs
Outcome	Frame broke again; did not fly at all

Iteration 3

Bracing, cap, first flight

Quadcopter frame, iteration 3

Added arm stilts and a clasping arm mount to stop the arms oscillating while the propellers spun. A top cap went on to protect the flight stack in an upside-down landing, and the leg stilts were shortened, since v2 was unnecessarily tall. I also added proper Raspberry Pi and battery mounts, each placed to match the drone’s weight distribution.

New: arm stilts + clasping arm mount (kills arm oscillation under prop load)
New: arm claw (reinforces the arm-to-body joint, the exact spot that broke on v1 and v2)
New: top cap (protects the flight stack if it lands inverted)
Shorter leg stilts; dedicated Pi and battery mounts placed for balance

Problem targeted	Arm oscillation, joint failure, weight balance
Outcome	The first version that flew

Iteration 4

Lightweighting + modular joints

Quadcopter frame, iteration 4

This iteration was about weight. Triangular cut-outs through the top body and the arms dropped the frame weight significantly. The v3 arm stilts had broken (their mounting was too thin), so I replaced some screwed joints with a friction-fit cylinder-in-hole design, which also saved weight by cutting hardware. Added a printed “Titan v4” logo on the top plate and a recessed groove to seat the radio.

Triangular lightening cut-outs through the top body and arms
Friction-fit cylinder joints replace some screws (lighter, tool-free)
New: recessed radio groove; printed “Titan v4” logo
Printed in multiple colours from separate parts

Focus	Weight reduction
Outcome	Flew well until the push-pins failed; hot-glue held them together for continued testing

Iteration 5 · current

Final build + autonomy

Quadcopter frame, iteration 5

The final build. I added a dedicated GPS mount and dropped v4’s friction-fit joints (the cylinders were too thin and kept snapping) and went back to screws. The print settings changed too: infill from 15% to 35% with a gyroid pattern for maximum strength. This version never broke, and it flew under the autonomous tracking code.

New: GPS mount
Friction-fit joints reverted to screws (v4’s cylinders were too thin)
Print: 15% → 35% infill, gyroid pattern
Printed tracking target (150 mm sphere)