Dexterity is definitely a problem. It goes beyond tasks in the field..even past harvesting, how do we grade them, wash them..pack them. So far only greens can be packaged robotically. I think earthbound farms have a robotic packing system for lettuce/spring mix. However another company tried packing peppers in a clamshell and just the intuitive way human packers will slide, squish, rotator and pack peppers was such a big deal because that's not something you figure with sensors or or grippers or vision system or algorithms. Because nature doesn't produce replicas of standardized sizes of sweet peppers.
What is required is to change the shape or size of the clamshell to suit robot pepper packers. That's a shift. Many such shifts in design are required.
I think that's a long long way away. For myself (and there are so many different tasks that can automated in farming), I have picked just seven field tasks that are repetitive manual tasks in small acreage farms that are in the area of soil prep, seeding and weeding.
Small farms don't enjoy economies of scale that large commodity farms do..how you grow corn is entirely different from how eggplants are grown. Everything we have automated and standardized in vegetable farming is compromised from the natural way to grow things.
There is a minimum acreage that justifies existing automation wrt costs. It's economies of scale. How do you grow 15 acres of eggplants with machines? If you can't deploy an army of hand weeders: 1. You can spray pesticides(cost=toxic) 2. You can select hybrid seeds that are resistant to certain diseases(cost=eliminates or drastically reduces bio diversity) 3. You can just export it(cost=loss of food security and sovereignty)
only small farms can produce vegetables. Mega farms produce commodities.
Building a tractor sized bot with sensors, vision system, gps, lidar is massively expensive. Too much data to aggregate. Just not viable. Not worth it. An acre of eggplants will yield about $6000 worth of eggplants. What is the point of spending million plus on an autonomous tractor? But.. A $30k machine that swarms an acre of two/day..even just to check when it's ready to pick, do a bit of weeding, monitor water and soil..do automatic seeding and transplanting...spray compost tea or spread compost is totally worth it. The key is that it does it by itself. Which means that when I am tweaking my website and selling tomatoes from my stand and chatting with my customers, my Eggplant bot is doing all the work. By itself.
I can watch from my screen. And if malfunction occurs, I would maybe lose one or two rows of eggplants. As long as the rest of the swarm functions, work gets done. But when the sole autonomous tractor breaks down, everything is lost.
That's 2k/Year per acre. Wages: One single employee would cost..at minimum wage..$15 x 40hrs/week x 4 weeks/month x 8 months = $19200. You'd need more than 2 workers to tend to 15 acres of vegetable farm. Even if the robot has a working life of 2 years, it's worth it.
Plus: human labour is better used elsewhere. It's kinder to human beings. 3. Machines learn and collect data. 4. Better health for soil, air and human beings. 5. Bots can work more than 8 hours a day and for more than 40 hours/week. They can work in the dark, cold and heat. They work efficiently. We save on $ on water, fertilizer etc that would otherwise be wasted on non-precision ag methods.
How is the costs on farm bots NOT being justified?
Building a 300lb human scale robot that does human tasks is relatively easy. Deploying a swarm of robots over 100 acres is relatively easier than sending one mega tonne tractor to do the work over 100 acres.
I think there might be some tasks that can never be replicated by machines. But it is viable if they are small machines. Industry is working on large machines because they are trying to model it like self driving cars. Designs come from existing farm
Machine designs without innovating. Because farming by itself doesn't seem to deserve tech disruption. Which..again..baffles me.
And you are absolutely right. Getting something to do something as simple as weeding would be a 2 year project for me if I engage and exploit and stand on the shoulders of giants. Remember that you can only field test once a year. Because soil changes with seasons. Weather changes every day. Which is why advances in this field must have started 5 years ago. Why it's stalling now and is compared with self driving vehicles is a mystery to me. The two are nothing alike. For starters..farm bot needs to go on a straight line..no opposing traffic..high torque low speed..low power. Trying to overlap the challenges of an autonomous car on something that is so much simpler in scope and trying to match them re costs doesn't make sense to me. The biggest pitfall imo is that they are trying to make massive tractor sized bots..that's the biggest stumbling block. Even if they come up with something miraculous like that, no one is going to be able to afford it. Frankly, I think it's daft.
Let me attempt an example:
I was weeding a row of violas a few days ago. Over fall/winter, it had grown super weedy. It takes me 90 mts to weed a 100 ft long row that is about 3 ft wide. There are three rows of violas and are staggered. That's about 300 violas plants and covered with all manners of weeds. This is what I do if I were an intelligent farm weeder:
1. I detect violas and weeds.
2. I gather Viola plant.
3. I search for the roots of the weeds.
4. I pull the weeds out.
5. I check for slugs or pests on violas.
6. I check soil.
7. I fluff soil.
8. I tamp the soil down.
9 repeat around next viola plant. Do it again 299 times.
This is what a robotic weeding crawler should do. It has a picture of viola plant. It slices the tops of the roots of everything that it can't recognize as violas of Viola leaves.
Or
A robotic rodent can be let loose in a row of violas and it will 'chomp' down everything that it doesn't recognize as a viola plant.
A tractor can never weed a row of 50 rows of violas. I would look to nature. In nature, gophers and bunnies and small critters destroy the plants. Machine solutions for farms should be based in nature.
I want a robot rodent to eat my weeds. I will let loose a dozen of them over 50 rows. They will be done in 2-3 hours.
I don't want a million dollar self driving fancy tractor mounted AI system to kill the weeds between a row of violas. It will not and can never do the job.
When you work soil, it pushes back. Every square feet of earth has resistance that works against you. Thats why farm machines weigh zillion tonnes. But a human being weighing 150lbs takes a hoe to a row and does it by hand without compacting soil. But he has to do it for hours and hours. It's slave labour.
Sometimes less is more. And in farming, this is certainly true. Nature should be an inspiration for farm bots. Not machines.
Do you still think farm robots have to be expensive and complicated? It's very helpful to me to hear other perspectives because it means someone who isn't entrenched in the kind of work I do sees it with a diff set of eyes. Many thanks and Regards.