Skip to content
Laser weeder · 01 light
Teardowns · 8 machines, part by part

How real robots are built, and why the hard part starts after the first one works.

Pick a machine. See how it senses, thinks and acts, what changes when you build a fleet of it, and which parts of the software Device to Cloud Flywheel does for you. Step through it with the buttons below, or keep scrolling.

Early accessDevice to Cloud Flywheel: free to self-host for your whole organization, or hosted by us.Get early access
0 · You may have seen this

You may have seen machines crawling through vegetable fields, killing weeds with lasers instead of herbicide or hand crews.

A young weed can look almost the same as the crop, and the machine has to find it and hit a growing point a few millimetres across while it keeps moving.

0 → 1 · How it works: sense, think, act

For every small step the machine takes, it runs these 5 steps in a loop.

Sense
Light, See
Think
Decide, Aim
Act
Fire
01 of 05, on the machineSense

Light

Inside the closed hood, LED bars light the bed the same way at noon and at dusk.

02 of 05, on the machineSense

See

The downward camera beside one laser takes a frame of the bed.

03 of 05, on the machineThink

Decide

The computer on the deck runs the weed model on the frame, marks the weed, and finds its growing point, the few millimetres where a shot kills the plant.

04 of 05, on the machineThink

Aim

The computer predicts where the weed will be when it reaches the laser row and turns the steering mirror in one laser module to that spot.

05 of 05, on the machineAct

Fire

With the hood interlock showing the hood closed, the laser fires and the mirror puts the beam on the growing point.

The software at 0 → 1

You build the hardware and its sense, think, act loop on makeshift software.

ssh weeder-01 'tail -f /var/log/detector.log'
scp weeder-01:/data/run-0412/*.mp4 ./to-label/
rsync model-v7.onnx weeder-01:/opt/models/   # hope nothing is mid-row

This works for one machine. It does not work for a fleet.

1 → 1000+ · From one to a fleet

Then you redo the software.

Going from one robot to two, let alone a thousand, means redoing much of the software: fleet management, firmware control, and a great deal of integration. It takes real money, forces hard choices about where to spend it, and raises the barrier to entry.

06 of 09, across the fleet

Log

The machine records video of the bed in short segments, whether or not anyone is watching.

Device to Cloud Flywheel can record the camera in short video segments, kept on the machine until they upload.

07 of 09, across the fleet

Return

When the link allows, the machine uploads its recorded segments.

Device to Cloud Flywheel uploads them when the link allows, and reports which programs are running.

08 of 09, across the fleet

Learn

The fleet's records, this weed among them, are the raw material for the next model.

Device to Cloud Flywheel moves the frames your team picks to your trainer, checked by checksum, when you start the job.

09 of 09, across the fleet

Ship

The new model goes out as a declaration of what each machine should run.

Device to Cloud Flywheel sends the new model to a first group, then you advance, pause or roll back.

↺ · Where Device to Cloud Flywheel makes the difference

What your team builds, and what Device to Cloud Flywheel does for you

Then the loop starts again at 01 light, with the new model.

You keep building
  • The weed model
  • Laser, mirror and timing control
  • The hood interlock and safety chain
You don't build
  • Connects each weeder with one command, so nobody has to SSH into machines in the field
  • Records camera video in short segments and uploads it when the link allows
  • Moves the frames your team picks to your trainer when you start the job
  • Sends the new model to a first group of machines, with pause and rollback

Get early accessSee every part of the laser weeder, drawn step by step

1 / 15

What you don't have to build

Around every machine sits the same software: connecting it, watching it, moving its data, and updating it safely. Device to Cloud Flywheel already does this, so your team spends its time on the machine.

  • Connect each machine with one command, without SSHWorks today
  • Declare what each machine should run, and see when one drifts from itWorks today
  • Run commands on any machine from one placeWorks today
  • Health telemetry from every machine, and an event logWorks today
  • Bring field data home as files, checked by checksumWorks today
  • Hand recorded data to training, and bring the new model backPartly
  • Roll out a new model or setting in stages, with pause and rollbackWorks today
  • Take over a machine remotely, with a deadman cut-off on the machineWorks today
  • Users with read, write and admin roles, and an audit log of who did whatWorks today

Device to Cloud Flywheel runs this loop for your machines.

Robot makers and model builders work on the same platform, so you spend your effort on great hardware and great models and leave the infrastructure behind.

  1. 01OnboardOne command connects the device.
  2. 02ConfigureSet its tunable settings, per machine.
  3. 03OperateSend it commands, and get telemetry and results back.
  4. 04Retire or repairSpot drift, fix it, or remove it.

Everything is a device, and every device goes through the same four stages.

RobotRuns beside the robot’s own software, for example LeRobot on an SO-101 arm.Works in simulation
TrainerTrains a policy from the runs your robots recorded.Partly
SimulatorRuns MuJoCo and Isaac Sim simulations.Works today
Inference serverServes a model to the machines that need it.Coming

Open source

Free, permanently

Self-host it for your whole organization. Unlimited devices and users, on infrastructure you control. We are in early access. The code opens to everyone at launch.

Hosted

Alpha

We run it for you. We add teams in small groups.

Enterprise

By conversation

It runs inside your own network, and we support it.

Already on ROS? Device to Cloud Flywheel runs your ROS stack as a managed workload. Also from us, coming: Device to Cloud Engineer, for the device-specific code your robot needs.

What robotics teams ask when they get stuck.

How do I deploy, monitor, troubleshoot and update robots once they’re out there?
Eng lead, scaling fleet1 → 1000+
I’m a hardware guy who put together a robot. What next?
Founder, first robot0 → 1
Should I use Isaac Sim? I’m not sure real2sim2real works for us.
Robotics engineer0 → 1
We need to test small ideas, but with our team size everything gets done one at a time.
CTO, early-stage agtech0 → 1
I’d buy first, then fine-tune as we grow.
CTO, funded agtech1 → 1000+

Paraphrased from conversations with founders, CTOs and robotics communities.

Get early access to Device to Cloud Flywheel.