The First AI-Based Robot Weed Killer | Hacker News Reader

The First AI-Based Robot Weed Killer

carbonrobotics.com

34 pointsjkuria19 comments

The title is false. This robot weeder uses AI and has existed for years. httpss://tertill.com/

Is it even possible to create a robot weeder without AI? It needs some sort of reasoning system to identify what is a plant.

> Is it even possible to create a robot weeder without AI?

No, if you use the AI communities definition "AI is any mathematical algorithm that makes a decision"

In the normal world -

I want a robot that can mow. Cut everything to the ground. I'll get a 10 year old to program the path for it (Or put in any beacons it wants)

Too expensive.

It's not AI, it's the price of equipment. (At this stage, the AI bit's the easy section)

Came here to say this too. Saw Tensorfield (2018) demo'd in Shenzhen a few years ago, nice people. Quick web search reveals ecoRobotix (2018; Switzerland), Blue River Technology (2021; Indiana), Farmwise (2019; California), Naio (France; 2018), HortiBot (2008; Denmark)... there are surely more.

Looking forward to when the 5 ton robot decides that humans are weeds

it does have an ED209 feel to it https://www.zonecollectibles.com/wp-content/uploads/2013/05/...

I am not a farmer, so this may be a really dumb question. Is this tool useful only in the weeks after planting?

At some point in the growing process, your cameras and lasers are not going to have a direct line of sight to the weeds. Your crops are going to block the view to the ground below.

Maybe killing the weeds before your crops grow too large is the only time you need to worry about it. Like I said, I am not a farmer.

I was surprised to not see it in a field where plants were actually growing beyond a few inches. I think you always care about weeds - they take water and nutrients which are especially important when the crop is coming in, not just the first week.

AI-Based = git repo that does simple image matching. How hard is find a weeds on dirt background? AI-Based is probably worth another 20 million when raising funds.

But on a serious note, this is great! Reducing chemicals and fertilizers is great for the environment and improving the quality of the food chain.

I'm sure that Monsanto isn't happy about this.

The hard part is deciding what is a wanted plant and what is a weed.

Yes, the example image shows onions, which seem like an easy match vs. broad-leaf weeds, but plenty of other vegetables would be harder.

450w total (150wx3) of laser power seems like it could start a fire. I doubt farmers would be allowed to use this in summer or early fall here in New Zealand.

I am sure they have thought of this already and don't run the lasers long...but still, running 3x high powered laser cutters outdoors unsupervised seems risky.

A 150w laser can cut 1.2mm thick steel.

The best option would be using femtosecond lasers that ablate rather than burn the material. No idea if that is possible at those power levels.

That would definitely reduce/eliminate the fire risk but I think the challenge with that approach is that you would effectively just be cutting the weed off at the surface.

With the CO2 laser they are basically cooking it to death. The long wavelength likely penetrates much further into the root as well, increasing the likelihood that its permanent. That said they could moderate the dwell time of the laser to avoid having things burst into flame if/where that is an issue.

I do wonder if they could use microwave instead.

Crop fields will be watered, not a big fire issue.

I'm guessing you haven't seen many fires.

Looks amazing. I wonder why it is so big though. The lasers and computers do not seem to be that large.

With "Cummins-powered drivetrain" and "runs all day and all night", my naive assumption is fuel storage.

That was my initial thought as well. Digging around the site you can see some images of the interior, it looks like they optimized for field serviceability and extensibility. There's also quite a bit of real-estate required for the powertrain, as it needs both a hydrostatic system and probably 10-20KW of electrical generation to drive all of the lasers and compute.

Some stats here: https://carbonrobotics.com/features

5 tons, 8 lasers, 75hp, 5 mph, 75 gallons of diesel fuel.

I'm a bit skeptical of the small wheels, though. Yes it is four wheel drive but I have a four wheel drive compact utility tractor that's about half that weight and the front tires (about the same size) are almost useless when the ground is wet. Unless this thing is really good at avoiding mud I think it would get stuck quite a bit (at least in Ohio).

Apart from the other reasons listed I’d guess to also make it look “farming” grade.

Reply on news.ycombinator.com