The FPV Drone
A perception-driven quadrotor: a printed frame, a real flight controller for stabilization, and a brain for the policy. Fly the twin first, in real rigid-body physics.
Drive it in real physics.
The four-part build starts with this: the design as a real MuJoCo twin you can pilot right now on Play. Take the controls, then build the body that runs it.
Fly the twin
A real-physics quadrotor. Throttle, yaw, pitch, and roll on any controller, through a nose camera or a cinematic chase view, the flight envelope you'll tune on the real frame.
Any controller: throttle, yaw, pitch, and roll, through a nose camera or a chase view.
▶ Drive the twin on Play ↗Four parts, one loop.
Body, brain, driver, policy. The design forces no single board, only that whatever you pick can take the commands and run the policy.
A printed quad frame
A printed X-frame sized to 3–5" props, with motor mounts, a board deck, and a camera cage. TPU feet for the hard landings you'll have while learning.
A flight controller, plus a mind
Stabilization belongs on a real flight controller; the policy belongs on a brain beside it. Pick the brain below, the flight controller is the constant.
Setpoints → flight controller
The policy sends attitude and throttle setpoints to a Betaflight / PX4 flight controller over serial. The controller holds a loop rate no laptop can; your brain sets intent.
Perception → setpoints
Fly line-of-sight to start, then offload vision to the companion board. The same setpoint stream drives the sim and the real quad.
Ways to power the same body.
Cheapest to heaviest. Every one drives the identical printed body and runs the same policy.
Flight controller + ESP32-S3 companion
~$40The flight controller (Betaflight / PX4) holds stabilization; the companion offloads perception or relays setpoints.
Raspberry Pi 5 + AI Camera
$150+On-board vision for autonomy, riding alongside the flight controller.
NVIDIA Jetson Orin Nano
$249A real vision-action policy in the air, within the weight and power budget.
Full specs and roles for every board: Digital Brains →
Open robots like this one.
The Maker Atlas maps real, open Physical-AI robots to each of our builds. These are kin to this one, open a profile to trace its boards, sensors, and source.
Crazyflie + AI-deck
Bitcraze's open 27 g nano-quad with an AI-deck (GAP8 RISC-V + Himax camera) running on-board CNNs, the nano-drone perception standard.
In the wildArduPilot
The open autopilot for DIY drones/rovers on printed frames + a Pixhawk FC and a companion Jetson/Pi for vision. GPLv3.
In the wildCogniFly
An open foldable, crash-resilient <250 g AI drone, a printed flexible exoskeleton + Pi Zero + AIY Vision, Betaflight. Open; prints flat then folds.
In the wildPULP-DroNet
A deep-learning collision-avoidance CNN running fully on the Crazyflie AI-deck's GAP8 (35–102 mW, ~139 fps), the seminal open nano-drone perception project. Apache-2.0.
See how they all connect: the Maker Atlas →
Live now: the twin. In progress: the physical unit.
The sim twin on this page is real and live today. The printable body and the driver are specced against the sim-to-real registry; the first physical unit is being assembled, and the STL pack and step-by-step land here the moment it's verified on real hardware, not before.
Design its body, program its brain.
Take this build into the Studio: model and export its printed parts on the CAD kernel, and write the MicroPython that runs on its board. Both open as full-screen workspaces.