PDA

View Full Version : Most Precise (Hobby-grade) Arm?



davezula
10-30-2013, 07:33 PM
Hello All!

Thanks for stopping by :)

I was looking to open a short discussion on the comparison of "hobby-grade" robotic arms such as the Trossen WidowX/Lynxmotion AL5D, etc... in terms of precision accuracy, speed, lifetime hours, overall build, etc. when compared to something like the Fanuc LR Mate.

Does there exist a relatively decent manufacturer making "high enough hobby grade" arms to break into low-entry level industrial use? Or would one typically use something like a WidowX for prototyping and require the purchase of a more professional arm for production...

Also among all of the entry level hobby-grade arms, which then might be considered the most reliable in terms of repetitive precision accuracy, speed and lifetime hours? Average to light stress loads.

Kind Regards and Thanks in Advance,
-Dave

Th232
10-30-2013, 08:06 PM
Does there exist a relatively decent manufacturer making "high enough hobby grade" arms to break into low-entry level industrial use? Or would one typically use something like a WidowX for prototyping and require the purchase of a more professional arm for production...


I suspect you'd need to define the tasks first. For example, a WidowX might be a reasonable substitute or test case for moving things on/off a conveyor belt, but if the task was placing electronic components then I would personally just jump straight to the Fanuc for reasons of accuracy and precision. Generally though my understanding is that a lot of things are done in simulation, e.g. ABB's RobotStudio. Playing with their gear was a real pleasure at uni.

Ditto for your other question of entry level hobby grade arms, what exactly is an "average to light stress load"?

davezula
10-31-2013, 04:50 AM
Thank you for the quick reply!

RE: tasks, my intentions were to have it in a more precise application - though not quite on the level of SMT components assembly.
I would aim to replace the gripper end with a pneumatic vacuum nozzle and use the arm to pick and place small ceramic tile surfaces each probably weighing < 1g. The precision tolerance in this scenario could probably work decently at +- 1 to 2mm.

In this sense the stress load would be more so on the repetitive motions of the arm ~50,000 repetitions at 180 degrees. Can the Dynamixel servos handle these workloads with continuous precision? Arm jitter is another variable - but I don't believe that relevant in this application (it's just impressive when you see something like the Fanuc move with such precision.)

I was told that the MX-64's rather than the AX-12's from Dynamixel for example can be programmed to deliver fairly decent accuracy... yet I'm still not sure what makes an arm "precise". Is it just a perfect combination construction material / servos / programming, etc?

Will definitely check out ABB RobotStudio - though I'm afraid it will leave me with the desire to go buy a high-end arm without the financial means to do so... :)

jwatte
10-31-2013, 11:38 AM
The arc angle resolution of the MX servos is 0.088 degrees (1/4096 of a full turn.)
However, the play / backlash in the gear train is greater than that, so at a minimum, you'd have to load it with a counter-spring, which significantly reduces your effective load carrying capacity.

Let's say you want to reach out 50 cm. sin(0.088 degrees) is 0.00154. Thus, the best precision you can get is 0.768 millimeters. Unfortunately, though, the flex in the construction (and backlash, as noted above,) will probably make this worse.

For making ceramic mosaic, I've seen people use stepper motor based X/Y gantries (with timing belts/pulleys) to great effect. Building those is also pretty simple and robust.

davezula
10-31-2013, 07:01 PM
RE: X/Y gantry, that might be a better way to go - certainly more affordable I think. Previously I had seen a similar application using an arm so I thought maybe that would be the way to do it... but it's probably overkill.

Some final questions then: does all of the above imply that something *like* an LR Mate could be fabricated at home with similar performance by purchasing the right servos + optimizing the build? Any recommendations for high-torque / high-resolution actuators outside the Dynamixel line?

Ebay starts to look very good with the various industrial servos on auction, but I guess that enters a whole new realm of interfacing - where something like the MX-64's could nicely plug into an arduino without too much headache.

Th232
10-31-2013, 07:05 PM
Your main issues will be rigidity and backlash as jwatte mentioned previously. Regarding rigidity in particular, there's a reason that industrial robots have such large/heavy arms. This then leads to a requirement for even more powerful servos. I suspect that by the time you're done making one you'll be running into a similar cost as that of buying the LR Mate, if not more.

Outside the Dynamixel line you've also got the Dynamixel Pros in case you haven't seen them. What's your budget for all of this?

jwatte
10-31-2013, 08:50 PM
If you're building ceramic mosaic, and you can provide some counter-tension to the servos in the arm, and you can wait for the flex in the assembly to settle before releasing each tile, and the reach is sufficient for your needs, then something like a Widow X might work for you. If you want to try using an arm, instead of getting some belts, pulleys, linear bearings, and stepper motors, then it could be worth a try (assuming you're not spending your pension savings to get one, because it may fail...)

However, I would be astounded if it actually could do 500,000 cycles without significant degradation in performance or failure. (Someone please correct me if I'm wrong!) Thing is, you can buy 10 new Widow X arms for the price of one LR Mate :-) (But the load carrying capacity is different, right?)
If all you want is the cheapest, most robust way to make tons of ceramic mosaics, then an XY gantry with linear bearings is the way to get there.

davezula
11-01-2013, 11:00 AM
@Th232 - RE: Budget, I'm trying to stay as low as possible, naturally. My max is around $400 for prototyping, and if things progressed smoothly I would probably not hesitate to invest considerably more into better gear. Though you're right on the rapidly increasing costs :S Will try and follow these suggestions and see where they get me.

@jwatte - RE: 500,000 (or even 50,000??) cycles I would definitely be impressed as well! Though to be fair I have no idea what the average lifetime on any of these hobby-level servos are. The AX/RX/MX lines (and some of the higher-end HiTec's) seem fairly robust... still it would be cool to see some performance data on these somewhere.

Also seeing as (indeed) what I'm looking for is "the cheapest, most robust way to make tons of ceramic mosaics" I believe I'll take your advice and go with the X/Y gantry! I'll start a project page soon enough if anyone else is interested...

CasperH
12-08-2013, 02:57 AM
I agree with above, a cnc gantry like system with something lick a sucker as end effector seems the best way to go.

Currently I am thinking you are looking for something like this:

http://www.hizook.com/blog/2012/08/02/artaic-revolutionizing-tile-mosaics-through-robotic-assembly#not-robotics


http://www.youtube.com/watch?v=7eQ1WceKTxA

Given your budget, maybe you can combine the following things:

http://www.amazon.com/Build-Your-Machine-Technology-Action/dp/1430224894/ref=pd_bxgy_b_img_z
http://www.amazon.com/Zen-Toolworks-CNC-Carving-Machine/dp/B002ARTLUG/ref=sr_1_3?ie=UTF8&qid=1386492617&sr=8-3&keywords=cnc+hobby+router

Assuming, you want/can do a some things yourself. I typed "cnc hobby router" in amazon for above search results.

What I find the tricky part to be of your project/idea is a decent delivery method for all the mosaic pieces (color specific?) and coming up with the CNC code to pick and place them. Have you looked in to the software? Can you show us an example of a "finished" piece when it is assembled by the robot?