Meet DEXI‑3.
Every student codes.
A Hover Lab subscription gives every DEXI‑3 student the full coding pathway (blocks, Python, missions, computer vision) running in the simulator from day one. Pick the kit for when that code flies on the real drone.
Grade 6 and up. That is a floor, not a guideline, DEXI‑3 is a real aircraft students build and fly. Younger classes fly the same missions in the browser with Ground School, then move to the drone.
The full coding stack, Raspberry Pi CM5, camera, and the open DEXI OS. Python, ROS 2, and vision from day one.
What makes it different.
Open the box, charge the battery, walk into a gym or classroom, and fly. A downward optical-flow sensor and a time-of-flight range finder feed PX4's EKF, so the drone holds its position over a tile floor, a carpet, or a desk, the same way every day, in every room.
Built to be opened
Every layer of firmware and software is public on GitHub, read it, fork it, ship your own.
Indoor first
No GPS required, optical flow + ToF give rock-solid hover over any indoor surface.
PX4 autopilot
PX4 1.17, the same open-source stack used for autonomous-flight research worldwide.
Real Linux
The Raspberry Pi CM5 + ROS 2 means you can deploy any code you can write, to a real drone.
Classroom-scale
A 3″ carbon fiber airframe sized for indoor learning, small enough for a gym, tough enough for a class set.
From blocks to code
The same curriculum spine as Ground School and Hover Lab, students grow into the drone, not out of it.
The drone Hover Lab
was designed for.
DEXI‑3 is the first aircraft in the DroneBlocks flight path and it stays in the fleet at every level above, the same block-coding foundation students start with in Ground School carries through to a real, programmable drone they assembled themselves.
Ground School
Block coding + browser simulator. No hardware required.
DEXI‑3
An aircraft they build, Python and ROS 2, flown indoors, GPS-free.
DEXI‑5 / DEXI‑10
AI vision, applied research, and industry applications.
You are looking at the on-ramp. Everything a student learns here carries up the flight path, the same airframe, the same code, the same curriculum spine.
Two doors. Same student.
DEXI‑3 ships in two SKU forms. The Flight Kit gets students building and flying. The Developer Kit adds the full coding stack: companion computer, camera, and the open-source DEXI OS. The Code Upgrade Pack bridges them when budgets are staged across years.
| What's in each | Flight Kit | Developer Kit |
|---|---|---|
| Airframe & propulsion | ||
| Frame | 3D Printed (CF upgrade) | Carbon Fiber |
| 4680 KV motors · tri-blade props | ✓ | ✓ |
| Flight stack | ||
| DroneBlocks H743 FC + ESC (PX4 1.17) | ✓ | ✓ |
| Optical flow + ToF range sensor | ✓ | ✓ |
| 3S battery, charger & ELRS transmitter + RX | ✓ | ✓ |
| Code stack | ||
| Raspberry Pi CM5 + Carrier Board | , | ✓ |
| Pi Camera Module 3 + programmable LED strip | , | ✓ |
| microSD with DEXI OS | , | ✓ |
| Capability | ||
| Manual flight + indoor optical-flow hover | ✓ | ✓ |
| DroneBlocks & Python coding in the simulator | ✓ | ✓ |
| ROS 2 + custom code on the drone | , | ✓ |
| AprilTag + YOLO computer vision | , | ✓ |
Built for how your
budget works.
Choose the Flight Kit if…
- Budget is staged across multiple fiscal years
- The class wants stick-time first, code later
- It's a fleet for an after-school flying club, not a robotics course
- The school is piloting drones for the first time and wants a low-risk entry
- Makerspace use case, the airframe is the project, not the autonomy
Choose the Developer Kit if…
- The course is Hover Lab or any coding-forward unit
- One purchase order is preferred over multi-year staging
- Robotics, AI, and vision are part of the syllabus
- Teachers are already trained on Python or ROS 2
- Research / college-prep program with custom-code expectations
The drone is the airframe.
The Code Kit is the platform.
A Flight Kit student finishes the term a confident indoor pilot. A Developer Kit student finishes shipping Python missions, landing on AprilTags, and writing their own ROS 2 nodes, on the same drone.
Hover Lab on my.droneblocks.io
A lesson-by-lesson Python pathway built for DEXI‑3: the block-coding bridge from Ground School, mission scripting, and vision challenges. Every Hover Lab subscriber gets the full pathway in the simulator from day one; the Code Kit is what flies it on the real drone.
AprilTag + YOLO out of the box
Pi Camera Module 3 plus the open-source dexi_yolo package, YOLOv8-nano object detection and AprilTag tracking running natively on the CM5. Precision landing, follow-the-tag, and AI-vision missions ship as reference demos.
ROS 2 + custom code deployment
A full ROS 2 Jazzy stack on the CM5 with a MicroDDS bridge to PX4: the same pattern professional autonomy teams use. Deploy your own packages, neural networks, and Node-RED flows over Wi-Fi.
Code, test, fly, in the browser
The DEXI Simulator lets students write and run the exact same code before it ever touches hardware, so a class can start on day one and the real drone is just where the code lands.
Same Python script.
Sim today. Real DEXI tomorrow.
No drone. No toolchain. No install.
Students open code-server in their browser, write a MAVSDK Python script, and fly an autonomous mission in the Unity sim. No drone. No toolchain. No install.
One line of config different.
The exact same file runs against a DEXI‑3 on the classroom Wi-Fi. Same takeoff, same waypoints, same land, just one line of config different.
address = "udpin://0.0.0.0:14540" # sim
address = "udpout://192.168.x.x:14540" # real DEXI
The whole bridge is one line. That promise is what lets a Hover Lab subscription deliver value the day a teacher buys it, weeks before the drone arrives, while batteries charge, when one drone has to serve thirty students. The simulator is not a toy version of the curriculum; it is the curriculum.
Hover Lab subscription +
Code Kit hardware
The hardware SKUs above are only half the offering. The full path is the Code Kit drone plus a Hover Lab subscription on my.droneblocks.io: one PO, one curriculum spine, students starting in the simulator on day one and flying the drone the day it arrives.
Sim → Code → FlyNothing about DEXI‑3
is a black box.
The drone software, OS image, ROS 2 bring-up, and 3D-printable parts are public. Fork them, modify them, contribute back.
Pre-built Raspberry Pi OS image for the CM5, bundled with the full ROS 2 stack and PX4 link.
github.com/DroneBlocks/dexi-osROS 2 launch files and bring-up packages, the entry point for offboard control and custom nodes.
github.com/DroneBlocks/dexi_bringupSTL files for replacement and customization parts, the same parts shipped on the production drone.
github.com/DroneBlocks/dexi-3d-printsThe fine print, on demand
Open any panel for the engineering detail: the software stack, how the layers talk, the full bill of materials, and the complete hardware spec.
Open by design, the software stack
DEXI OS is a Raspberry Pi OS build pre-loaded with the autopilot link, ROS 2 bridge, and DroneBlocks bring-up. No proprietary cloud, no closed SDK, no hidden firmware, every layer is inspectable and forkable.
- Python, first-class, the same language as Ground School & Hover Lab.
- ROS 2: the full stack on the CM5: topics, services, and actions.
- PX4 1.17, on the integrated H743 FC. Tune params, modify modules, build custom firmware.
- Custom Code: ship your own packages, neural networks, mission scripts, and Node-RED over SSH/Wi-Fi.
How the layers talk
| PX4 ↔ ROS 2 | MicroDDS bridge, PX4 uORB topics published into the ROS 2 graph on the CM5. |
|---|---|
| PX4 ↔ QGroundControl | MAVLink Router on the CM5. |
| Pilot ↔ Drone | ELRS 2.4 GHz, Nano RX. |
| Developer ↔ Drone | Wi-Fi to the CM5, SSH, ROS 2 networking, file deploy. |
Hardware bill of materials
| Frame | 3″ carbon fiber (Developer Kit); 3D-printed on the Flight Kit, carbon fiber upgrade available |
|---|---|
| Motors | 4680 KV brushless ×4 |
| Propellers | 3018 tri-blade |
| Flight Controller / ESC | DroneBlocks H743-based FC w/ integrated ESC, PX4 1.17 |
| Companion | Raspberry Pi CM5 on DroneBlocks custom carrier board |
| Flow & Distance | Optical flow + downward ToF range finder |
| Camera | Raspberry Pi Camera Module 3 |
| Battery | 3S 11.1 V LiPo, XT30 |
| Flight time | 10+ min typical |
| TX / RX | ELRS TX + ELRS Nano RX |
| Storage | 64 GB microSD |
| LEDs | Addressable strip |
Complete technical specifications
| Airframe & propulsion | |
|---|---|
| Frame | 3″ carbon fiber (Developer Kit); 3D-printed on the Flight Kit, carbon fiber upgrade available |
| Motors | 4680 KV brushless ×4 |
| Propellers | 3018 tri-blade |
| Flight time | 10+ min typical |
| Mounts | M2.5 standoffs, 3.0 mm |
| Flight controller & autopilot | |
|---|---|
| FC / ESC | DroneBlocks H743, integrated ESC |
| Firmware | PX4 1.17, custom DroneBlocks branch |
| Optical flow | Downward |
| ToF range finder | Downward |
| Companion computer | |
|---|---|
| Module | Raspberry Pi CM5 |
| Carrier | DroneBlocks Pi Carrier Board |
| Storage | eMMC + microSD |
| Wireless | Wi-Fi + Bluetooth |
| Carrier I/O | |
|---|---|
| Power | 7–28 V → 5 V @ 4 A buck (TPSM86638) |
| Camera | 1× MIPI CSI 22-pin |
| I²C | 2× JST-GH, 4.7 kΩ pull-ups, ESD |
| UART | 2× JST-GH, ESD |
| Addressable LED | GPIO JST-GH |
| Ethernet | 10/100 JST-GH |
| USB-C | USB 2.0 |
| Display | 1× Micro HDMI |
| Fan | 4-pin JST-SH, PWM + tach |
| Camera | |
|---|---|
| Sensor | IMX708, 12 MP |
| Default ROS 2 stream | 640×480 @ 30 FPS MJPEG via camera_ros/libcamera |
| Configurable | Up to full sensor resolution |
| Vision & AI | |
|---|---|
| Object detection | YOLOv8-nano via dexi_yolo |
| Inference | ONNX Runtime, CPU inference on CM5 |
| Fiducials | AprilTag CPU detector |
| Radio | |
|---|---|
| Transmitter | ELRS handheld TX, RadioMaster-class |
| Receiver | ELRS Nano RX |
| Band | 2.4 GHz |
| Power | |
|---|---|
| Battery | 3S 11.1 V LiPo, XT30 |
| Charger | Classroom-safe |
| Software stack | |
|---|---|
| OS | DEXI OS |
| Middleware | ROS 2 Jazzy |
| Autopilot | PX4 1.17 |
| Programming | DroneBlocks, Python, ROS 2, custom |
| Ground station | QGroundControl over MAVLink Router |
Where the code gets tested.
Teams fly this aircraft at the Advanced Vertical Robotics Competition: an autonomous course, a payload, and a clock. The curriculum on this site is what gets a team ready for it.