Explore the V3

Get to know every part.

Turn the robot, zoom in, and select a component to see how it works.

Loading the 3D model…

Drag to rotate · Scroll or pinch to zoom · Click a part

Select a part on the robot or from the list.

V3 CAD design, revision A08 · Assembly and field testing in progress

Engineering log

Three subsystems, one field robot.

We're building CropScout as three subsystems that bolt onto one base: a tracked undercarriage, an ML/AI crop-vision stack, and a precision sprayer. This page is the evidence: the working MVP, the boards we designed, the CAD, and the plan behind it.

Watch the MVP demo Try Vision

Subsystems, one base

$3,000

Funding received so far

Custom PCBs in the MVP

Farmers interviewed

Subsystems

What we're building

Each subsystem is owned by a founder, with its own hardware and its own test. They meet at one mechanical interface, one 24V bus, and one CAN network.

01CAD frozen · fabrication in progress

Undercarriage

Tracked drivetrain sized for greenhouse rows and wet soil: track carriers, idlers, drive sprockets, split road rollers, hubs and axles on a welded frame.

02Live model · experimental screening

ML / AI

Crop and leaf detection plus disease screening, running as ONNX models in the browser today and on the robot's own compute next.

03In CAD · parts ordered

Sprayer

Precision spraying on the same base: a 5 US-gallon tank, a 24V pump, and solenoid valves that dose per plant.

The MVP

Working prototype, demoed in public

A real, driving, sensing rover: a steerable four-wheel base, a controllable front arm, soil and air sensing, and a wireless controller talking to it over MQTT. Built and presented at the ASU Fulton Innovation Showcase in front of 500+ people.

CropScout V2 undercarriage, a CAD render: tracked base with the spray tank and camera mast — The SolidWorks assembly, fully defined: battery tray, 24V 20Ah battery, 5-gallon tank, tracks, idlers, sprockets and road rollers.

01CAD frozen · fabrication in progress

The field-scale base, in CAD

The MVP was a four-wheel rover the size of a suitcase. Farmers told us to make it bigger and let it carry more than one tool, so the V2 base is a tracked undercarriage designed from scratch in SolidWorks.

VTR056 rubber tracks on 100 mm idlers and drive sprockets, four split road rollers per side.
Low centre of gravity: the tank and battery sit inside the frame, not on top of it.
Every attachment mounts to the same top plate — the sprayer first, a weeding module next.
Motor nodes move from UART to a CAN bus, on a Teensy 4.1 controller.

A CropScout Vision result: a tomato leaf screened as septoria leaf spot, 88% condition score — A real result from the live tool, in the browser.

02Live model · experimental screening

Disease detection you can try

CropScout Vision runs two stages: first, is this a supported crop, and a leaf or a fruit; then the closest covered condition for that crop and part, with a confidence score and the alternatives it considered. Models are exported to ONNX and run in the browser with ONNX Runtime Web, so photos never leave the phone. We judge every model change against our own field photographs, not a lab validation split — that is where studio-heavy public datasets fall over.

Crop check first: a wrong guess about the crop is never allowed to become a confident diagnosis.
Condition screen: the closest covered condition, a score, and what else it considered.
“Was this right?” under every result — the field feedback the public datasets can't give us.

Open CropScout Vision

CropScout V2 sprayer module, a CAD render: the tank, pump shelf and hose routing — Concept model — the interfaces the sprayer, camera mast and probes share on one base.

03In CAD · parts ordered

Precision spraying, per plant

Spraying was the second-largest time cost and the single largest cash cost our farmers named. The sprayer module is what CropScout does first with the base and the vision stack.

5 US-gallon tank in a cradle with hold-down bars, filled and rinsed through the top cap.
24V diaphragm pump on its own shelf, fed from the same pack as the drivetrain.
Solenoid valves and a nozzle boom switched per nozzle, so a flagged plant is dosed and its healthy neighbour is not.
Dose, pressure and litres per row, logged to the farmer's phone with the rest of the telemetry.

Electronics

Our own PCB stack

The motor-driver PCB, assembled and tested: an ESP32-S3 and four H-bridge driver chips — Assembled and tested.

The same board in KiCad: a two-layer layout with four H-bridge channels — The layout in KiCad.

Every PCB in the MVP was designed and assembled by us: schematic, layout, fabrication, bring-up and debugging. The board pictured is the motor-driver node — an ESP32-S3 driving four TLE9201SG H-bridges over SPI, taking 5–40V in and regulating to 3.3V and 5.5V on board. It's one of seven PCBs in the rover's distributed architecture; firmware is MicroPython, with MQTT over TLS for commands and telemetry and a UART daisy chain to the other boards.

Next revision: the main controller moves to a Teensy 4.1, for more headroom, CAN bus and real-time control, and a Jetson-class module joins the stack so Vision runs onboard.

Schematic, layout & firmware on GitHub Motor-node PCB datasheet Watch the bench demo

Open build days

Come build with us

Two open build days at ASU Polytechnic's SIM Building, for anyone who wants to see the robot come together.

ML / AI subsystem build day
Friday, 9 October 2026
CropScout assembly
Thursday, 30 October 2026

Funding & links

Where the money and the work live