intermediate robot arms?

Try giving more specifics about your requirments and we prbably can offer some potential solutions.

1 What is your " Budget " ? $ 300 , $ 3,000 , $ 100 ?

1. What is your accuracy requirements.
2. What is your "ulitmiate goal for the arm" i.e lift a quart of beer , or twist off the bottle top
3. What is your time budget and engineering experience ?

if you have $300 then get a Scorbot on Ebay its a very capable 1 kg lift arm with controller, if you lurk the Scorbot III listings you'll find one every couple of months and they usually go for $300 -400 each.

If you have less money but more time then you can build and arm from aluminum and metal. To get the 1 kgs lift you can not use RC hobby servos, instead you will want to use surlplus Pittman motors and gearboxes. Attache these to a pully system and you will get your 1 kg lift.

For $3,000 you should be a ble to get a Puma or Milicron Arm that will lift a few Kg and control it with great accuracy.

What can you tell us ?

Thanks much for the response!

Under $1k would be great, but if $2k or so made a difference I can probably go there. $3k is probably over the top unless it really was the perfect solution (hardware and software library-wise).

I can probably work within a range of accuracies. However, this isn't an area of robotics I've read up on so I'm not sure what to expect. If i just imagined what would be most useful, I'd say I'd like to be able to get the end-effector to a desired position within 5-10 mm. But like I said, if that's too high of an expectation, I can work with something coarser.

The arm will be painting. It needs to be able to lift/select a brush, dip it in paint, and apply that paint. Ideally I'd be able to make brush strokes (place, drag, and lift) as opposed to just dabs, but I'm flexible.

I have time, but not much engineering experience. I'm good with my hands, and build things out of wood, but haven't built any electronics. That said, I have decent programming experience (rusty, but usable), have a strong understanding of 3D computer-aided design, and understand computers very well. I expect this to be a long-term project, but am also anticipating a hefty programming requirement (see below). As such, I had hoped to buy a complete arm instead of starting from scratch.

Another requirement is that I'll want to be able to programmatically and dynamically control the arm, hopefully using a language such as python, ruby, or java. I expect I'll need inverse 3D kinematics so that I can direct the end-effector to specific locations. Ideally, the arm would include software libraries that supported that as my math chops aren't up to that task.

Thanks for these suggestions, i'll look into them. It could be that 1 kg is more than I need... 0.5 kg might do it. I'll have to do some weighing.

If I were to go the build-it-myself route, are there 3rd party control libraries out there I could use to control it, and to hopefully do the inverse kinematics? Also, how much money would be saved going this route?

thanks again!

On Jul 27, 6:29 am, snipped-for-privacy@dodgit.com wrote: [...]

What kind of painting is it to do? Paint a house? A wall? A window frame? Or paint a picture?

Part of the solution might be a special purpose effector rather than a hand. Can you use spray painting for example instead of a brush?

-- jc

Painting a picture on supports ranging from 12" to 24" square.

Nope, needs to be a brush. And ideally, I need to be able to pickup different brushes--one brush won't do it.

If that's what you need, look into robotic tool changers:

Each tool (brush) has a standard collar attached to it, and the arm has a powered tool changer which can grab and release those collars reliably.

John Nagle

Then may I suggest not a robotic arm but a printer plotter machine?

Instead of a pen you use a brush.

Are you wanting to use tubes of paint or pots of paint?

There are probably electric paint dispenses (or maybe use electric liquid soap dispensers to store the paint to squirt out different amounts to be mixed.

-- jc

I thought of that. However, I'd also need to be able to control the height of the brush. I know that I've seen rapid prototyping hardware that works with a basic plotter type mechanism (say, ejecting material in a bead) onto a platform that moves up and down. In this respect, a plotter type (I suppose this would be a 'Cartesian' robot?) might work.

I am also after a certain visual aspect to the device, and the arm would be a lot more interesting to watch than the plotter. But I'm not ruling it out. Would that kind of device be a lot less expensive? Easier to build and/or buy?

I'm considering supporting different kinds of paint. One would originate in tubes. Another would have to be picked up from small pots onto the brush. Your idea would work for the former.

Thanks for the link. This idea makes a lot of sense and I hadn't thought of it.

A cartesian/gantry robot would be much easier to control than the classic articulated arm you're thinking about. If the painting doesn't need to be vertical, then a SCARA arm would be another good configuration to consider.

With any of these, I suspect the main control difficulty will be maintaining the correct distance/pressure between the brush and the painting. That's why a cartesian or SCARA arm might be better than a fully articulated one. Cartesian is nice because painting coordinates directly map to joint coordinates... Inaccuracies in the SCARA control might give a more human effect, though. Since the main joints in a SCARA arm are horizontal, they can be powered by smaller motors...

- Daniel

Yes a LOT less expensive and easier to accurately control.

You can always incorporate the "human effect" mentioned by Daniel with simple software.

I have seen electronic magazines with these setups as do it yourself projects. Often for automatic cutting of materials etc.

-- jc

Back when us CAD designers used pen plotters, it was common for other people in the company to stop at the plotter and watch it do its work. I always said I should charge admission.

Some earlier models of plotters had stationary pen holders... those may be more adaptable to your goals.

I know you want to do a brush, but I suspect that a sprayer(pehaps airbrush like) would be more dependable. There is an unpredictability when ever you handle something "soft"... anything material that doesn't behave exactly the same way each time.

An idea I've had in the past (but never really persued) would be to make a 3-D gantry-type milling machine. Have it carve a 3-d sclpture from foam, and then paint it using the spray head. This way it may be possible to re-create something like a vase with a painted pattern. The same tool path used to machine away the final layer of foam could drive the painter head... just tell the CAD program that the "tool" head is a little bit beyond the end of the spray head.

is a good source for a relatively low-cost gantry mill. '4x'4x6" version is $7,500

One web site with a forum and lots of infor for a do-it-yourself gantry mill is

Joe Dunfee

If you really want to emulate human painting, industrial robotics hardware is up to the job. A tool changer for brush changing, a force sensing wrist (see "

"), and a standard low load robotic arm will do the job.

You could probably use the Lynxmotion hobbyist robot arms if you could add decent force feedback. You want to servo the brush into about the right position, then use force feedback in the direction perpendicular to the canvas during painting to bring the brush into contact. Then maintain that force level during the brush stroke. Break contact with a positional move. One axis of force feedback would be enough for this.

The hard part is that hobbyist-grade software usually isn't capable of things like positional control in two dimensions with force control in the third, out at the end of a jointed arm. And hobbyist robots tend to come with crappy sensors. But the technology is well understood. It's used industrially for grinding, drilling, sanding, and polishing applications. References are available.

Once you have force feedback working, you can use it to dip the brushes into jars of soft paint in a controlled way.

Further up the planning hierarchy, you can get paint-by-number software which will take in a picture and generate a paint-by-number plan. Then you need a stroke planner, and sequencing for picking up paint, cleaning the brushes, and brush changing.

John Nagle

I have seen commercial robotic arms draw pictures of people with great speed and dexterity. As a hobby robotic arm they are too expensive but if the arm is making money for you then that is a whole other ball game.

The hobby robotic arm i played with was too slow, weak and required you add all your own sensors. I could make my own sensors but the slowness and weakness took all the fun out of it.

-- JC

John:

Do you have any references that you would recommend? I'm more interested in how the sensors work rather than a mathematical treatise on inverse kinematics.

-Wayne

One of the more common devices in industrial robotics is a 6 degree of freedom force sensing wrist. This detects positional and rotational forces in each of three axes. The wrist is a slightly flexible part equipped with strain gauges, usually with some onboard electronics to bring out the amplified signals in a useful form. This is overkill for many applications, but it's generic, so it's a common solution industrially.

Simple force sensors are widely available, and not too expensive. The general idea is that you bond a strain gauge to something with a bit of springyness.

See

for a $9.90 force sensor to play with. Range is 0 to 3Kg, and it's easy to interface to an A/D port, so that's about what you'd need to measure the force on a paintbrush.

Academic robotics went to force sensing in the 1980s, and by the 1990s, industrial robots tended to have force sensing. The general idea is that you use preprogrammed positioning to get to about the right place, then fine-tune using force sensing.

I did some of this stuff using industrial robots at Stanford back in the 1980s, when the technology just barely worked. Today the parts are much better.

John Nagle

Thanks for the link. Those look like easy to use force sensors -- far easier than trying to measure force via motor stall current.

Is there any comparable easy to use rotational force sensor? I'm thinking of something with two aligned axes sticking into a blob of some sort and some wires sticking out of the middle.

-Wayne

Look up "torque sensor".

What you usually get is something that looks like a shaft collar with electrical connections. You can get bare strain gauges and wrap them spirally around a shaft to make a torque sensor.

See

for a torque sensing system used for winding rubber band airplanes.

John Nagle

Thanks again for all the discussion. I've done more looking into cartesian setups, and can definitely see how this would be easier than an arm. I also noticed that some of them seem to not require support on two sides (e.g. the Y is only supported by the X side), which makes watching the process easier. Certainly the addition of torque sensors and tool changers would improve it as well.

Given that, can you suggest some good starting points for building and/ or buying a cartesian robot? The shopbot link leads to a variety of cartesian bots made for CNC. Is a CNC setup the best route to pursue for prebuilt, or will someone out there prebuild general purpose cartesian bots (or which I can design and/or purchase my own end- effectors)? I'd like something rugged and quick, but still within budget.

If not prebuilt, where would I look to to find some of the structural components for such a machine? Would I have to design all of these myself and have them machined?

I will also start looking up old electronics magazines to see what I can find project-wise.

Finally, is there a generic software library for controlling such a robot?

thanks again

One might also consider an electronic "touch" sensor to detect initial contact with the surface, then use additional incremental movements to increase the desired brush "squish" on the surface.

The tone of your post is you don't know much electronics? I think you would need to find a complete project to copy if that is the case and those questions should be part of it.

There is probably some stuff on the internet?

Do you do any programming if so what language?

A project should come with explanations as to how to read sensors or send movement commands to the stepper motors but all the rest would be up to your imagination and programming ability.