Gallery, Projects and General > Project Logs

Remote contrtolled manipulator to eradicate common wild oat

<< < (2/3) > >>

RotarySMP:
Great project concept.

Why the human? Given the cost of labour northern europe, completly eliminating the need for labour in this process must be an attractive goal.  What about splitting the spotting and pulling into two separate datalinked machines. The spotting sounds like the perfect application for a flying drone, which could quickly survey the field and pinpoint the wild oats, meaning the walker can save time manoeuvring directly to the suspect, rather than surveying a whole field.

How are you thinking about doing the machine identification? Is there some obvious difference such a reflectivity a certain spectrum, or are you going to need a neural networks based recognition software and do some training for it to recognise leaf features?

How hard is it to pull these weeds? When I think of the weeds in my rose garden, there are some with strong stalks and weak roots which are easy to pull, and then there are the barstards with weak stalks and strong roots which just break off and grow straight back.

Will this walker be tethered to the drivers vehicle for electrical power, or autonomous. You would need a lot of power to do a whole field on autonomous battery power.
Mark

eskoilola:
Why the human?
As much as I like the idea of having an autonomous machine left alone in the field outrooting the weeds I really hate the idea of machines completely replacing humans. Maybe I am a communist?
The best neural network by far is sitting right between our ears. I have been working with neural networks, genetic algorithms and whatnot with my daily job. These have limitations and require a slow and painful learning which in this case might become quite overwhelming because of the diversity on the tasks.

How are you thinking about doing the machine identification?
I assume You mean the regognition of the bad guys here. Obviously it has to do with some sort of optical regognition. I am pretty sure that the weeds have some optical properties that the actual crops do not have. For example how do they react on UV light or other parts of the light spectrum? The result could be superimposed on top of video feed to make the weeds stand out.

What about splitting the spotting and pulling into two separate datalinked machines.
The walker itself must be linked to a server with a data link. Analyzing the entire field in advance with a drone or similar is tempting as this would help in finding the weeds as the viewing angle is different. This also means that it should be possible to clearly identify the weeds. What it also means is accurate navigation. Is it possible to do this without accurate navigation?

How hard is it to pull these weeds?
It depends on the weed and the soil. The wild oat has to be pulled all the way out - roots and all. If my memory serves me correctly these are not that hard to pull out but the top of the plant is quite weak so it has to be gripped near ground. It could be a pair of rolls for example. Sort of a mangle.

You would need a lot of power to do a whole field
Correct. Therefore some sort of a generator must be on board. Battery is definitely not an option. The construction should be such that almost any generator with suitable AC output can be used.

CrazyModder:
Crazy project, I like it.  :beer:

On the power aspect. Don't forget that it is easy to multiplicate machines. I would not rule out batteries just yet. You could have a non-moving generator at the edge of the field (hence, no limits on size/fuel), and then use more bots than you normally would. Some of them will be charging while others will be working, and there will be a constant exchange. Obviously you have to get your calculator out to check if it is really feasible and take into account how well the batteries will hold up (basically, the e-car problem). But this may make the moving parts so much smaller, lighter and hence cheaper that it may just be worth it.

eskoilola:
I have been looking around internet for hydraulic joint. Did not find anything that would be suitable.
Those that I found were capable to lift a car or broke a leg with ease. This "animal" does not need that sort of strength.

I also thought that the hydraulic actuators might be a little bit leaky. If the fluid volume is big enough and if there are bending sensors on each joint then leakiness is no more a problem. If these can be made without soft sliding parts - the better.

Hrere is an illustration of my thoughts about this thing. Has anyone made or seen that kind of hydraulic actuators ?

eskoilola:
After some search I finally found the english name of this type of actuator. It is called Rotary Vane Actuator. And they are really expensive  :coffee:.

I will try to fabricate at least one when I get back home from this darn assignment. It does not seem to be too challenging to do as it is symmetric and round. Precision required - yes.

I had earlier a Russsian made mini lathe called M2. With that one I learned to do things with this type of equipment. The result: at least 50 kilograms of scrap iron and maybe 3 kilograms of good stuff.

I actually found a picture of this lathe. It has a round bed and is supposed to be quite accurate. I can tell: It is nothing but accurate.


The lathe I have now has not been used at all - it is brand new. I have to clean, oil, level and calibrate it. I also purchased my very first micrometer  :dremel:.

What I really learned with that russsian lathe was grinding HSS. I seldomly use carbide as it is initially so dull. Carbide also seems to make a better finish if the cut is deeper so with carbide one has to plan ahead. Sharpening a HSS tool is just 20 wipes with the Norton oilstone (grey side) and there You are. Razor sharp. Do that with carbide.

Edit: There is an entire page dedicated to this exotic lathe.
http://www.lathes.co.uk/russian/

Navigation

[0] Message Index

[#] Next page

[*] Previous page