The Biogrow ground robot straddling soybean rows at dawn, firing a red laser pulse at a weed between the plants.

Autonomous laser weeding · Anderson, South Carolina

The drone maps.
The robot fires.

Biogrow Robotics removes weeds with a near-infrared laser instead of herbicide. A survey drone finds and georeferences every weed before the machine enters the field — so the ground robot never has to decide anything in real time. It just drives to a coordinate and fires.

808–980 nm

NIR diode laser, thermal ablation

GNSS RTK

Centimetre-level field positioning

Zero

Herbicide, tillage and operators

The problem

Chemistry is running out of road.

Weed control in row-crop agriculture still rests almost entirely on broad-spectrum herbicides. That model is failing from three directions at once.

Resistance
Herbicide-resistant weed species keep spreading
Regulation
Tightening chemical restrictions in the U.S. and Europe
Cost
Rising input prices with diminishing efficacy
Demand
Buyer and consumer pressure for residue-free produce

Mechanical alternatives disturb the soil. Flame systems are energy-hungry. Existing robotic weeders still either spray a chemical or try to identify every weed in real time from a moving vehicle — which is where their speed and reliability go.


The architecture

Sensing and execution are separate machines.

This is the whole idea, and it is what our patent filings claim. Every other laser weeder tries to see and shoot from the same moving platform. We split the job in two — and the ground robot gets dramatically simpler, faster and more reliable as a result.

FLIGHT DAY BEFORE DEPLOYMENT FIELD RUN UAS survey LIDAR + multispectral 3D point cloud of the canopy CNN separates crop from weed Each weed georeferenced by GPS → detection accuracy target > 95 % Weed dataset Coordinates, canopy size, density Distance to the crop row Path optimised into a traverse Transferred over an encrypted link → the robot leaves already knowing UGV execution GNSS RTK, centimetre-level Cross-references position vs list Laser fires only on arrival No real-time recognition required → 808–980 nm, 2–10 J per pulse The two platforms never need to be in the field at the same time. Mapping can happen days before the run. This separation is the subject of the company's provisional patent filings.

Swipe the diagram sideways to follow all three stages.

A LIDAR survey drone flying above a soybean field while the ground robot works the rows below.
STEP 01

Survey

A LIDAR and multispectral drone flies the field and builds a 3D point cloud, separating crop from weed by height, reflectance and geometry. Every weed becomes a georeferenced record: position, canopy size, distance to the crop row.

STEP 02

Plan

The weed dataset is handed to the ground robot over an encrypted link before deployment. A path optimiser turns thousands of scattered coordinates into the shortest safe traverse of the row structure.

STEP 03

Execute

The robot drives the rows on RTK positioning, cross-referencing its own location against the list. When it reaches a coordinate, the laser head aims and fires — thermal ablation of the weed tissue, no contact with soil or crop.

Why this matters

A robot that already knows where every weed is doesn't need to be clever in the field.

Real-time detection is the hardest part of a weeding robot, and the part that breaks first — in dust, in low sun, under a partially occluded canopy. Moving that work into the air, ahead of time, removes it from the critical path.

Less onboard compute Higher traverse speed Robust to dust and light Lower unit cost
Close-up of the laser optics head under the robot chassis, firing a focused red beam at a weed seedling.

Thermal ablation, not burning

A focused near-infrared pulse heats the weed's growing point until the cells fail. Nothing touches the soil. Nothing drifts onto the crop.


The robot

Built to be manufactured, not admired.

The platform exists to carry three things into a crop row and keep them working: the optics, the power and the positioning. Everything else was pushed off the vehicle on purpose.

Three-quarter studio view of the Biogrow four-wheel high-clearance autonomous weeding robot.

Design visualisation. The first functional prototype is the objective of the current programme.

Platform

Chassis and drive

Footprint
≈ 6 ft × 5 ft
Configuration
Four-wheel, high clearance
Drive
Electric, independent wheel motors
Energy
Rechargeable battery pack
Terrain
Adapted for uneven ground
Operator
None — no tractor, no cab, no driver

Payload

Optics and control

Emitter
High-power NIR diode
Wavelength
808–980 nm
Pulse energy
2–10 J, adaptive
Optics
Adjustable focal length, servo steering
Navigation
GNSS RTK, IMU, local vision
Compute
Embedded control core, PC + MCU stack

> 95 %

Weed detection accuracy from the aerial survey pass.

Target · from the patent specification

≥ 95 %

Task success rate in Q3 field trials.

Target · programme KPI

> 4 h

Continuous operation without failure at prototype stage.

Target · programme KPI

On the cost and savings figures you may have seen

Estimates of input-cost reduction per hectare, and of unit economics against machines costing over a million dollars, come from our internal financial model. They depend on crop, weed pressure and current herbicide programme, and they have not yet been validated by a third party in the field. We will publish measured numbers when the field programme produces them, and not before.


Where it works

Row structure is the requirement. Not the crop.

The architecture works wherever plants are grown in defined lanes and the weeds sit between them. That covers most of the acreage in the world, and almost all of the acreage where herbicide is the current answer.

Wide view of a vast soybean field at sunrise with the robot working in the distance.

Cereals & oilseeds

Soybean, wheat, barley and corn. Wide acreage, predictable row geometry, and the crops where herbicide resistance is doing the most economic damage. This is where the first field programme will run.

The robot treating a weed between rows of leafy vegetables.

Vegetables & row crops

Lettuce, brassicas and similar bed-grown crops, where weeding is still largely done by hand and where a chemical residue is a commercial problem. Here the labour line is the cost being replaced.

The robot operating at night, lit by its own work lights, firing laser pulses at weeds.

Organic production

Certified operations where no herbicide is permitted and labour is the binding constraint. Thermal ablation touches neither the soil nor the seed bank, so it fits the rules by construction.


Company

An engineering company with one specific idea.

Not that lasers can kill weeds — that is established. Our claim is narrower and more useful: that a weeding robot becomes far cheaper, faster and more reliable the moment you stop asking it to recognise anything. That is a manufacturing problem, not a research problem.

The Biogrow Robotics team outside the company entrance.

The team

Engineering, agronomy and operations under one roof.

Engineers working on the robot's electronics and optics in the Biogrow engineering laboratory.

The laboratory

Where the control stack, the vision models and the laser optics are built and bench-tested.

Dr. Serji Armikhanian

CO-FOUNDER

Engineer and inventor with a background in complex systems, intellectual property and mechatronic product development. Leads the system architecture and the laser and optics work.

Gabriel Sakc

CO-FOUNDER

Military engineer with a track record in controlled-environment agriculture and mission-critical systems. Leads the agricultural application, the field programme and industrialisation.

Intellectual property

The separation is the patent.

Two provisional applications cover the system, both naming Dr. Serji Armikhanian and Gabriel Sakc as inventors. The independent claims turn on the same structural point: an aerial platform performs detection and georeferencing before deployment, and the ground platform executes weed elimination from that pre-acquired coordinate data without real-time recognition.

Filing 1
Autonomous robotic system for laser weed eradication with LIDAR-supported drone
Filing 2
Asynchronous aerial–ground laser weed eradication system
Status
Provisional

Programme status

Advanced design. Prototype next.

We would rather tell you this plainly than have you find it out later. Biogrow Robotics is a pre-prototype company. The architecture is designed, the bill of materials is complete, suppliers are selected, and the laser subsystem has been tested in the lab. The next twelve months build the working machine.

Q1 — FOUNDATION

Architecture freeze

Mechanical and electrical design final. Long-lead components delivered. Closed-loop motor control on the test rig.

Risk removed: technical feasibility

Q2 — INTEGRATION

Working prototype

Full assembly, sensors integrated, end-to-end control stack live, first autonomous task executed.

Risk removed: system integration

Q3 — VALIDATION

Field & reliability

Representative-environment testing, laser calibration, mechanical fatigue, pre-compliance safety and EMC.

Risk removed: operational reliability

Q4 — SCALE PREP

Production readiness

DFM review, cost-down roadmap, technical and QA documentation, pilot-ready unit.

Risk removed: productisation


Contact

Tell us what you grow, and where.

Every message reaches the founders directly. We answer enquiries within two business days, including the ones that turn out not to be a fit.

Direct contact

Office

1428 Pearman Dairy Road
Anderson, South Carolina
United States

Response time

Within two business days

Investors

Investor enquiries

We are raising the round that funds the prototype programme. Select Investment in the form and we will send the executive summary, the use of funds and the quarterly KPI set. The full technical package, including the patent figure set and the bill of materials, follows under NDA.

Send a message

We use what you send only to reply to you. No newsletter, no list, no third parties. If you would rather not use a form, write to info@biogrowrobotics.com.