Field robotics · Kelowna, BC

Robots that monitor and tend to every tree & vine. Today, that is a human-pushed sensor cart that records the rows it passes and ties each gopher mark to a place in the block, with irrigation issues and canopy health on the roadmap.

  • Designed and built in Kelowna, BC
  • Field runs in Okanagan orchards and vineyards
  • Every candidate finding checked by a reviewer

What we do

Row-by-row field data, checked by a person.

We collect it with a human-pushed sensor cart. Its cameras record each row it passes; early prototype models flag possible gopher mounds and sprinkler stakes, and a person confirms what goes in a report. A leaf scanner is in bench testing.

  • A cart ground camera frame with a prototype model box on a gopher mound
    Prototype modelModel output, checked by a person

    Gopher mounds

    Ground cameras record each row. An early model flags frames that may show mounds; a person confirms or rejects each one before it counts.

    Okanagan orchard, October 2026. Model output; prototype, not validated. Demo-only model, trained on footage that includes this pass.

  • Cart ground camera frame with prototype model boxes on irrigation stakes
    In developmentModel output, not yet checked

    Irrigation stakes

    An early stake model marks sprinkler stakes in cart camera footage. Growers have shown interest in stake checks.

    Okanagan orchard, October 2026. Model output; prototype, not validated; not yet checked by a person. Demo-only model, trained on footage that includes this pass.

  • Chart of one apple leaf's relative reflectance across 18 bands from 410 to 940 nanometres, read on the bench
    Bench prototype

    Leaf scanner

    Reads relative leaf reflectance across 18 bands, 410–940 nm. It is not yet calibrated against lab tests, so it gives no nutrient advice; pairing it with lab tests is planned for 2027.

    One apple leaf, bench reading, October 2026. Raw relative reflectance.

  1. 01Walk

    Push the cart down each row. Cameras, lidar and GPS record as it goes.

  2. 02Flag

    Early models flag frames worth a look. A person confirms or rejects each one before it counts.

  3. 03Report

    You get a map of reviewed marks by row and location, as leads to check on the ground.

01The problem

Gophers are hard to get ahead of on foot.

Most of the growers we interviewed told us they want gopher control. Today, scouting a block means walking it.

01 / 03

Found late

Gopher mounds and ground damage can be easy to miss until the trees or vines above them are already affected.

02 / 03

Hard to find again

A problem spotted mid-row may not be tied to a precise location, so follow-up can mean walking the row again.

03 / 03

Hard to compare

Without a consistent record, it can be hard to tell whether a block is getting better or worse over a season.

The sensor cart at the head of a vineyard row in fall, Okanagan, October 2026

02The platform

Walk the row once. Know what's in it.

An operator pushes the cart down the row at walking pace. It records the ground along the row, and every finding comes back tied to a place in the block.

  • Mark gopher activity

    Mounds are marked by the operator during the pass; a person reviews candidate findings before they go in a report.

  • Go straight to the problem

    Each finding is tied to a row and an approximate position, so a follow-up check can start at the spot.

  • Compare passes as the season goes on

    Findings are tied to row and position, so the same row can be compared between visits.

How a pass works1 of 4
An operator pushes the cart through an autumn orchard row

Workflow illustration

Technical details

We aim to add capabilities to the same platform over time.

Platform
Human-pushed cart, about 0.9 × 1.2 m and 30 kg
Row width
At least 1.5 m clear between canopy
Cameras
Colour ground cameras and monochrome navigation cameras. Near-infrared cameras are planned.
Lidar, GPS, IMU
Shape and location data so each finding maps back to a row and position
Power
LiPo battery pack today; a LiFePO4 pack is planned
In the block
The operator's phone provides the data link; no Wi-Fi or external power needed in the block

03What the cart sees

What the cameras are learning to find.

Our first models rank gopher-mound frames and test box detection. Grower reports stay human-confirmed. Irrigation and canopy work are on the roadmap.

A cart ground camera frame with prototype model boxes on gopher mounds
Model output, checked by a person
Prototype in training

Gopher mounds

Grower reports stay human-confirmed. Gopher detection is still a prototype; field accuracy has not been validated.

  • StatusPrototype detection
  • ModelIn training on our field data
  • In reportsReviewer-confirmed frames

Not yet claimed: detection accuracy, and time or cost saved per block. We'll publish these once they come from validated field data.

04What you get back

From a field visit to a report you can use.

Explore three ground scouting reports: row maps, operator marks and camera photos, brought together in one view.

These are anonymous copies of real field reports. Grower names, exact visit dates and geographic coordinates have been removed. The maps show the scanned rows on a 2018 public aerial photo, rotated and cropped to the scanned blocks. Findings are leads to check on the ground.

05Pilots

What a demo or pilot involves.

We offer free field demos to Okanagan growers this season, and a pilot can follow a demo. Here's what to expect.

  1. Intro call

    We talk through your blocks, what you're watching for, and which rows make sense to start with. We'll also ask what other pain points you have, and any ideas you have for the platform.

  2. Site visit

    We walk the agreed rows with the cart. We'll agree on which rows and how long with you up front, and schedule around your work.

  3. Review

    Every candidate finding is checked by a person, and anything we couldn't confirm is marked that way.

  4. Your report

    You get your block map and findings, and we go through them together. Early on every finding is reviewed by hand, so turnaround is agreed with you up front and may shorten as detection improves.

Cost

Field demos are free. If a pilot follows, we agree its terms with you in writing first.

Your data

You own the raw data collected on your land. We keep a copy to train our detection models, and delete it after training if you ask.

Disruption

The cart is pushed by hand down existing rows with at least 1.5 m clear and doesn't need to touch the trees or vines.

06Roadmap

From recording rows to working them.

We aim to add capabilities to the same platform over time.

  1. Record gopher activity for review

    Free field demos with Okanagan growers, with each finding checked by a reviewer.

  2. Ground and canopy in the same pass

    Planned: ground issues and canopy condition recorded alongside gopher activity, without a second walk.

  3. Leaf scanner

    A bench prototype reads relative leaf reflectance. It is not yet calibrated against lab tests, so it gives no nutrient advice; pairing it with lab tests is planned for 2027. Soil-probe timing depends on scanner progress and funding.

  4. Less walking, more doing

    Under-row weeder field testing is planned for 2028. Pruning follows on a robot of its own, with no date set.

07FAQ

Questions growers ask.

Does the cart touch or damage the trees?

The cart is pushed by hand down an existing row with at least 1.5 m clear and only takes pictures and measurements; nothing on it is meant to touch the trees, vines or fruit.

Do I need Wi-Fi or power in the block?

No. The cart runs on its own battery, and the operator's phone provides the data connection.

Will it slow down my crew?

The cart moves at walking pace down one row at a time, and visits are scheduled around your work.

What happens to my data?

You own the raw data collected on your land. We keep a copy to train our detection models, and delete it after training if you ask.

Where do you operate?

We're based in Kelowna and currently working with growers in the Okanagan.

Ryan Collins in an Okanagan vineyard

08About

Built in the Okanagan, guided by growers.

Iron Leaf Robotics Inc. is a British Columbia company founded in Kelowna by Ryan Collins, a solo founder and registered EIT with an electrical engineering degree (mechatronics option), production robotics experience at 4AG Robotics and aerospace electrical leadership at UBC Okanagan. Ryan has been full-time on Iron Leaf since October 20, 2026.

Ryan designs and builds the cart and tests it on local orchards and vineyards. That keeps iteration fast and keeps the work tied to problems growers actually have.

Our approach is to understand why the established answer exists before replacing it, and to only count what the field data supports.

Free field demo

See it on your rows.

Tell us a little about your operation and we'll get back to you to set up an intro call.