This is cool but I have to disagree with this phrase from the article.... "Given the scale of farming today, treating weeds chemically is really the only practical way for humans to keep them under control".

It's the easiest way to keep weeds under control but not the only practical way. (Source: I worked for the weed science department (yes, they had one at my school) while at university.)

Here are some other methods for reducing weeds:

1) Before planting, till and water the field causing weeds to germinate. Then harrow the field killing all the weeds. Do this multiple times if needed.

2) Good old fashioned cultivation as was done before the herbicide period. (The article on the robot was posted a few days ago and in response I posted the below... it has a lot of pictured of weed control devices which may be interesting to someone working on robotic weed control....https://archive.org/stream/SteelInTheField/Steel%20in%20the%...)

3) Crop spacing. Space crops so they out-compete the weeds (this is already usually done).

4) Keep fields clean. Keep the ditch-rows clean. Don't let weeds go to seed. After a period of time the bank of weed seeds in the soil is reduced and there are less weeds to control

I know herbicides are faster and cheaper (if we don't consider externalized costs). But they aren't the only "practical" way to eliminate weeds. It does take good management and more effort to eliminate or reduce herbicide use though. I personally think it's very worth it. Oh, and I'm excited about robotics. I don't know about this device, but I do think field robotics is the future.

I have countless weeks / months experience in dealing with manual weed removal. At times when I was younger I spent 14+ hours a day out in the field destroying weeds with a hand hoe.

One of the main crops I was doing it for was squash. The squash plant is pretty broad and the hand hoe allowed you to get close in under the plant. That's something this technique couldn't handle.

As someone working on minimum wage, I could get through a pretty large area in a day. Sometimes there would half a dozen of us in a gang. I'm not sure how much a device like this would eventually cost but I suspect for the same money you could buy a lot of minimum wage manual labour.

Either way, cool prototype!

The architecture of the ROS system is worth looking at if you haven't seen it. Everything has a URL and processes communicate using YAML messages by publishing or osubscribing to topics. (Its very simple but a bit more liberating than REST!)

http://wiki.ros.org/ROS/Tutorials/UnderstandingNodes

http://wiki.ros.org/ROS/Tutorials/UnderstandingTopics

http://wiki.ros.org/ROS/Tutorials/UnderstandingServicesParam...

I don't see it detecting any actual weeds? The detection is just targets and the demonstration of punching the weeds is stationary. Which leaves out the key challenge - distinguishing between weed & crop.

Parts of my family are in the bedding plant / greenhouse business. One cool technology is the transplanting machine. This transfers seedlings from "plug trays" to "flats" and replaces an assembly line of ladies doing this work. Here is one design:

http://www.hamilton-design.co.uk/transplanter.html

The ones they actually used were the "Harrison Transplanter".. could not find a video, but here is one http://bid.sheridanauctionservice.com/images/lot/3471/347193...

Harrison is neighbor of ours who worked at Brookhaven National Labs but then became an entrepreneur. He sold a bunch of them.

This is the kind of thing where robotics is really exciting. It does an end run around every problem with pesticide use by taking it out of the equation entirely.

Cool prototype.

...now make one that mounts on a front 3pt hitch and you might actually sell one.

No need for a special robot when a different piece of equipment with its own power source has to go down the field anyway

I was amused by "A Bosch Start-Up" on the side of the machine. Now everything is a start-up.

It's a good example of how play in the future will also go the way of jobs.

Why garden when your neighbors will have amazing gardens cheaply designed and looked after by robots.

And it'll be smart, it'll know exactly how many weeds to leave so it doesn't look to perfect, exactly when to plant and pollinate, order things online when needed.

That's brilliant.

What about using a laser to destroy the weeds?

So, HN, this looks like really fun and interesting stuff to work on. Any ideas on how a normal developer with a CS background can get into something like this? Either working for this company, or one like it... Or hobby projects that are within a normal person's budget.

Let me know if I should elaborate further and make an Ask HN out of this?

I think eventually farming will be completely robotic and we'll have farm on our roofs. Household waste will be processed and used for it. Robots will harvest crops for us and even get them to market if we aren't around to eat them, probably selling to people in the neighborhood.

How many years until a robot like this will cut your grass, weed the lawn, and weed your flowerbeds?

Looking at the video, they seem to be using a Microsoft Kinect sensor.

I agree, but to be fair, you could even skip that and just have a remote pilot looking at the rolling video feed and clicking on targets. It still beats having people actually walking under the sun and ruining their backs to pull out weed, or blanketing with herbicide.

I'm not so sure punching is a great solution though. Weed is incredibly good at reforming from very very small bits left standing.

The problem of automated plant identification has been examined [0, 1] with good results. In [0] the classification is performed on footage from an uncontrolled agricultural setting, and the authors achieved a correct classification rate of 65% on tomato plants. Given that modern industrial agriculture is monoculture you only need to positively detect one variety of plant in an area. To avoid crop destruction due to incorrect categorization the robot could be used as a first pass and restricted to punching plants that are classified with a high degree of confidence. This could still reduce the amount of human labor necessary to tend to the field, but it might not reduce it by a degree that makes the robot a sensible economic solution. It may fail to reduce the labor needed, but I do not think that is likely.

I wonder how the performance of plant recognition degrades over the course of the crop cycle, if at all. We see in the video that some leaf matter remains after the weed is driven into the ground, and this leftover matter will influence classification during the robot's next pass. Unless the leaf matter is removed between passes it will accumulate throughout the crop cycle.

In [0] the researchers note that one reason they chose to deal with seedlings is that there is relatively little plant leaf occlusion at that stage. In [1] an end user is relied on to take a photo with a light, untextured, background and does not at all deal with partial plant leaf occlusion. That paper is not applicable to the uncontrolled field scenario.

I also found [2] while writing this comment, but have yet to read it. It boasts even better results (>90% correct for corn, 73.1% correct for tomatoes in an uncontrolled setting)than [0]

[0] http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.42....

[1] http://neerajkumar.org/papers/nk_eccv2012_leafsnap.pdf

[2] http://www.mdpi.com/1424-8220/11/6/6270/pdf

I imagine if the robot is the one who planted the original crops it would know exactly where they should be. Any other growth is defined as a weed.