AI Robotics · Precision Weed Control

I built GreenBeam to treat the weed, not the whole lawn.

I'm Aryash Shyam, founder and inventor of GreenBeam. I started with a working prototype that combined computer vision and precision-light targeting, then kept building toward a connected autonomous system with a Robot, Garage, and App for residential lawns.

Part of my motivation came from researching the health questions and ecosystem effects associated with routine herbicide use. I wanted to explore a different engineering model: identify the individual plant first, then treat only the confirmed weed.
Back GreenBeam on Indiegogo ↗ Explore GreenBeam.us ↗
2025 3M Pennsylvania State Merit Working development system Engineering validation underway
GreenBeam autonomous precision weed-control robot
Why I built GreenBeam

One weed should not require treating an entire lawn.

Before I started building the robot, I researched what routine herbicide use can mean beyond the weed itself. A lawn is also where people spend time, pets run, rainwater moves, pollinators forage, and non-target plants grow. That made me question why broad-area chemical treatment is still the default when the actual problem may be a small number of individual plants.

My idea became simple: if technology can decide plant by plant, treatment should be plant by plant too.
01 · Human health questions

Cancer concern was part of my motivation.

The International Agency for Research on Cancer classifies glyphosate as probably carcinogenic to humans, Group 2A, and 2,4-D as possibly carcinogenic to humans, Group 2B. These are cancer-hazard classifications, not a statement that a particular labeled lawn exposure will cause cancer. The U.S. EPA has reached a different conclusion for glyphosate, saying its earlier assessment found no human-health risks of concern from labeled uses, and EPA is updating that assessment in 2026. I saw the disagreement itself as a reason to ask whether routine broadcast treatment can be reduced when a precise alternative is possible.

02 · Runoff and drift

What is sprayed does not always stay where it was sprayed.

EPA documents that herbicides can reach streams through stormwater runoff, spray drift, groundwater discharge, and atmospheric deposition. Once herbicides reach aquatic systems, effects can begin with plants and extend through habitat and food relationships that support fish and invertebrates.

03 · Non-target plants

A weed-control chemical can affect plants that were never the target.

EPA's current glyphosate review says ecological risks are primarily to non-target plants near application sites. More broadly, EPA notes that herbicide exposure can change aquatic plant communities and, in turn, alter habitat, food availability, community structure, and ecosystem function.

04 · The engineering response

I wanted the treatment footprint to shrink to the individual plant.

GreenBeam uses onboard computer vision to evaluate what it sees, distinguish a target from plants that should be left alone, and direct precision light only after a target is confirmed. The goal is not to make a better spray pattern. It is to remove routine herbicide spraying from the treatment process altogether.

That research became the design principle behind GreenBeam: see the plant, make the decision, treat only the confirmed weed.
Research behind this section: IARC: Glyphosate · IARC: 2,4-D · U.S. EPA: Herbicides in aquatic systems · U.S. EPA: Glyphosate review
The GreenBeam system

I kept expanding the prototype until the pieces could work as one system.

GreenBeam now brings together the Robot, Garage, and App. My goal is a connected workflow that lets a homeowner define where GreenBeam works, schedule a mission, let the Robot operate autonomously, return to charge, and review what happened.

GreenBeam robot rear three-quarter product view
01 · ROBOT

Perceive, navigate, decide, treat

The Robot combines onboard vision, AI compute, RTK positioning, LiDAR, environmental sensing, all-wheel-drive mobility, and precision-light treatment.

Computer vision Edge AI RTK LiDAR All-wheel drive
02 · GARAGE

Return, dock, recharge

The Garage gives GreenBeam a home base for automatic docking and charging through integrated roof solar or standard household power.

03 · APP

Map, schedule, monitor

The App is designed for digital boundaries, treatment zones, schedules, live status, treatment history, regional weed profiles, and system control.

What I engineered into GreenBeam

AI only matters here if the robot can understand the lawn around it.

Moving from a bench prototype to outdoor autonomy meant solving more than weed recognition. I had to think about where the Robot is, what is around it, how it moves, when it should stop, and how a plant-level decision becomes a precise physical action.

AI

Onboard computer vision

Plant imagery is evaluated on the Robot so GreenBeam can decide whether a plant is a target or should be left alone.

RTK

Precise positioning

RTK positioning supports location awareness and systematic coverage inside the digitally defined operating area.

3D

LiDAR and sensing

Environmental sensing helps the Robot understand obstacles and surrounding geometry as it moves.

4×

All-wheel-drive mobility

All-wheel drive is designed for the slopes, edges, paths, ruts, and variable ground found in real residential yards.

◎

Digital boundaries

The operating perimeter is defined in software instead of requiring a buried boundary wire.

✦

Precision-light targeting

After a target is confirmed, GreenBeam is designed to direct precision light toward the weed's growth region while limiting exposure to nearby non-target plant material.

The design principle stays the same across the whole technology stack: know where the Robot is, understand what it sees, and make the smallest treatment decision possible.
Explore the full technology stack ↗
GreenBeam App home screen
The GreenBeam App

I wanted the lawn to have a dashboard.

The App brings the physical system back to a simple homeowner experience. Define where GreenBeam works, create treatment zones, set a schedule, follow a mission, review treatment activity, and configure weed profiles relevant to the region.

01 Map the lawn. Define a digital operating perimeter without a buried wire.
02 Create zones. Separate front, back, side, or priority treatment areas.
03 Set schedules. Choose days, operating windows, and active zones.
04 Follow progress. See mission status, route, battery, and progress as GreenBeam works.
05 Review treatment history. Keep a record of detections, treatment locations, and previous missions.
06 Configure regional weed profiles. Prioritize weed profiles that are more relevant to the area where GreenBeam operates.
Explore the GreenBeam App ↗
GreenBeam Garage

I designed a home base so autonomy does not end when the battery gets low.

The Garage is intended to give GreenBeam a known place to return, align, dock, recharge, and get ready for the next mission. The charging design supports integrated roof solar and a standard household electrical connection.

01 · SOLAR Integrated roof solar
02 · OUTLET Standard household power
03 · WORKFLOW Return · Align · Dock · Recharge
GreenBeam Garage with integrated solar panels
From prototype to production path

I built the core idea first. Now the work is making the whole system repeatable.

GreenBeam began as my student-built prototype. It has grown into an integrated development system, and the current focus is structured engineering validation before production engineering and manufacturing preparation. I want measured results to replace marketing assumptions as testing progresses.

COMPLETED

Working prototype

I demonstrated the core computer-vision and precision-targeting concept in physical hardware.

INTEGRATED

Connected system

Robot, Garage, and App were brought together as one GreenBeam development platform.

CURRENT

Engineering validation

Defined protocols are evaluating vision, plant discrimination, mobility, positioning, boundaries, obstacle response, docking, weather behavior, and targeting.

NEXT

Production engineering

The next stage is refinement for durability, repeatability, manufacturability, serviceability, and product readiness.

GOAL

Customer delivery

Supplier, assembly, quality-control, and manufacturing work ultimately lead toward GreenBeam operating in real customer lawns.

See engineering validation ↗
The project behind the product

GreenBeam started as my science and engineering project. I kept going because the problem was bigger than the competition.

I began by researching the drawbacks of routine broad-area herbicide treatment and asking whether robotics could make weed control more selective. I assembled the early robotic hardware, wrote the Python-based vision software, integrated the targeting logic, and tested the first physical prototype.

That prototype earned Pennsylvania State Merit recognition in the 2025 3M Young Scientist Challenge. I could have stopped there. Instead, I kept developing the idea into a Robot, Garage, and App because I wanted to find out whether plant-level precision could become a practical homeowner system.

The question I am working on now is different from the one I started with: can GreenBeam become reliable, measurable, manufacturable, and simple enough to move from successful prototype to real product?

Read the GreenBeam story ↗ GreenBeam System ↗
GreenBeam is live on Indiegogo

Help me move plant-level weed control toward production.

Backing the campaign supports the next stage of engineering validation, product refinement, production engineering, manufacturing preparation, and the work required to move GreenBeam from a working development system toward customer delivery.

The idea that started GreenBeam still drives the product: treat the confirmed weed instead of routinely spraying the surrounding lawn.
Back GreenBeam on Indiegogo ↗ Visit GreenBeam.us ↗
Campaign pricing, packages, availability, and delivery information should always be confirmed on the official Indiegogo campaign page.
A few questions I get most often

GreenBeam, in plain English.

Why did herbicide use motivate me to build GreenBeam?

I was concerned by both the health questions and the environmental footprint of broad-area herbicide treatment. IARC classifies glyphosate as probably carcinogenic to humans and 2,4-D as possibly carcinogenic to humans, while EPA has reached a different conclusion for glyphosate when used according to its label and is updating its human-health assessment. EPA also documents runoff, spray drift, effects on non-target plants, and impacts that can propagate through aquatic ecosystems. That research pushed me toward a simple engineering question: if I can identify one weed, why should I treat the plants around it too? IARC · EPA

Is GreenBeam a robotic lawn mower?

No. GreenBeam does not mow grass. I built it for autonomous precision weed control. Its primary job is to identify target weeds and treat them individually.

Does GreenBeam use herbicides?

No. The system is being developed around precision-light treatment rather than routine blanket herbicide spraying across the lawn.

How does GreenBeam know where to operate?

The GreenBeam App is designed to define digital boundaries and treatment zones. RTK positioning and onboard sensing then support autonomous operation inside that mapped area.

What happens around obstacles?

GreenBeam combines LiDAR and onboard sensing for obstacle awareness. Obstacle response is one of the capabilities being evaluated through structured engineering validation.

Is GreenBeam already a finished consumer product?

Not yet. GreenBeam is an integrated development system moving through engineering validation. The next stages include production engineering and manufacturing preparation.

Where can I support the project?

GreenBeam is live on Indiegogo. I recommend using the official campaign page for current packages, pricing, availability, and campaign updates.

Read the full GreenBeam FAQ ↗