Any of you use NI LabVIEW for robot control?

[motion guru]

winfield - we have done quite a few labview applications with multiple motion platforms, data acquisition systems, tons of signal conditioning and synchronizing of data to fractions of a msec - it is great for the user whos lab is staffed with college grads who have been trained on it and can support it.

I would say that the population of robot users and general automation users who have used it is roughly 0.01% . . . the most popular graphical interface for programming on planet earth is ladder logic. Google IEC 61131 and see what you think - this is the basis for 95% of the controllers that are commonly used in industrial applications.

The greatest installed base that LabView has is Mindstorm Legos sets for kids - so this might change in 10 - 15 years.

[Winfield]

Originally Posted by motion guru

winfield - we have done quite a few labview applications with multiple motion platforms, data acquisition systems, tons of signal conditioning and synchronizing of data to fractions of a msec - it is great for the user whos lab is staffed with college grads who have been trained on it and can support it.

I cetainly didn't learn LabVIEW in college. Sherman:set the WABAC to college where I learned to program using 80 column punch cards in BlatIV Fortran on an IBM 360. Learned Basic, C, C++, VB, LabVIEW, and G-code on my own. BTW also self taught machinist as my background is EE. It doesn't take a college to be able to learn new stuff...but it helps.

[RalcoBe]

I am a Labview programmer, use it every day.
Mostly for machine User interface, datalogging, labo equipment, imaging, mostly under Windows.

There are big differences inbetween Labview:

-Windows
-CRIO
-RT
-Single board
-Vision
-Motion control

The design rule in our factory for machine control is still Omron PLC or Omron Motion + free GUI, so I won't move very quickly to RT.

.

[BadBeta]

Originally Posted by motion guru

First strike against it is that you need a PC to run it - a machine is automatically out of date within 3-5 years of installation if you are on a windows platform, hard drives fail due to vibration, boards become obsolete when you jump from ISA to PCI to network links to the PC.

You certainly have far better knowledge of robotics than me, but on this occasion I think you cut Labview a bit short.

While their are mainly known for windows based measuring tasks and such Labview has many more platforms than Windows. Among them real time computers/ OS, FPGAs and ARM microcontrollers.

Indeed I personally think this is their main issue; because all the different flawors have their own features, capabilities and limitations. It makes for a rather large and blurry mixture where everything is called Labview, even though they have to be approached differently. (And marketing wise everyone still thinks of Labview as Windows and measuring).

We use Labview for our prototypes which are basically robots. As far as reliability goes we have the entire program running on an FPGA chip - we don't even use the real-time computer portion of the Labview Compact Rio system. (No need for complicated movements - lots of IO though and different communication protocols).

That said it seems to me that Labview isn't a fully mature robotics control system just yet, and that the hardware prices are prohibitive for widespread use.

The lure of Solidworks integration and simulation is real (and back then one of the reasons we went for the system). But again, it doesn't feel mature to me just yet.

Overall though I have to say I'm quite impressed by the system, and I have no doubt they will become a very good platform sometime in the future. At least for prototypes - the hardware might remain too expensive for general use.

[RalcoBe]

Altrough I use it every day, I'm not completely pro.

Their succes is their biggest disadvantage.
After a 6-hours course, a normally-built-person "can" program Labview, "can" make a program run. Years after that, most of them still don't understand basic things like dataflow.

It's (for me) the easyest looking program, the quickest to make a jump start, but also the most complex at the end, when details give you sleepless nights.
It's also the easyest to make huge mistakes.

.

[bryan_machine]

The people having pain with old PCs clearly have fallen into all sorts of traps of the PC industry. The two killers are:

a. Special boards. Anytime you see a PC project that involves special hardware that is supposed to go into the PC, run away. All such special hardware needs to have super well documented interfaces and connect to the pc via some super strong long lived interface (ethernet...)

b. Very poor programming practices - signal timing based on processor loop counts, and other such things. The problem is that even high level experts have a hard time getting mid level experts to not do this. People who are expert at something entirely different will have a very hard time.

One *could* have a setup that made heavy use of *commodity* PCs that would be highly likely to be reliable 8 or 12 years out. But it *must* be designed to use very standard driver interfaces, very standard hardware. These are very different bets.

1. That you'll be able to get a particular speed processor and motherboard, with particular old slots, to run a card of which there are only 7 in the world, to speak some magic protocol.

2. That you'll be able to buy a PC that runs TCP/IP over ethernet....

But of course, getting this right is a hard task even for an expert, and strong forces in the PC industry are working against you. (Looked for a laptop with a serial port lately? No luck? That change was done for super good reasons, and I had a very small hand in it. Is a pain in the butt sometimes isn't it? Sorry about that...) This means there's margin and profit for Fanuc, Siemans, DT, etc. So they'll tend to be around.

The inverse to this is that the PC industry is truly truly huge, and if you can work with and keep up with commodity products, it's in some sense more likely to be available than any particular vendor of special controls. (Fanuc might go bankrupt. The entire PC industry won't.)

So the "just buy a controller from a reliable vendor who will support stuff for a required time, and hook that to a PC via ethernet when so desired" turns out to be a fine plan.

And my shop, my big CNC machines talk to PCs via either ethernet or USB stick. It'll be a while before PCs don't support those things.

(And in a business venture I'm involved with, that needs to reanimate some old hardware with this same problem - it's clear the solution is to make a box that speaks old-hardware-interface out one side and either USB or ethernet out the other. Ethernet is maybe a better choice.)

[RalcoBe]

I agree with most things you say.

Where I don't agree:
"a PC project that involves special hardware that is supposed to go into the PC, run away."
versus
"and hook that to a PC via ethernet when so desired"

You need an interface between the machine controller and a PC. Ethernet or PCI, stays all the same. They all have their (dis)advantage but you can't go without interface. Internal or external board, stays the same for me.

There are some nice FPGA solutions for an internal PCI board. It depends on what you want/are. Someone wh's running production is better off with a Fanuc/... controller, and the kind of (expensive but professional) service these companier deliver.
Other people doing development are better off with a flexible internal board.

2 combinations that don't work:
-a production guy with an old defective prototype machine
-a developer with an off-the-shelf professional controller, but undocumented and with closed design.

[bryan_machine]

RalcoBe - the issue is that the slot standards change. You can build a new ISA compliant board, but finding a new motherboard to insert it into is very difficult.

If that same board is in an external box that talks to the PC via ethernet (say) you are in better shape.

ISA and EISA have in effect gone away (hard to find new machines that support them.) Serial is slowly going away. Ethernet and USB are hyper commonly used (even more than serial was) and so they are good bets for long term support. [But in 1994, would anybody imagine a PC without a serial port?]

I agree about your 2 combinations that don't work....

In fact, there's a great line: "If a system works, but noone understands it, it is doomed. If a system does not work, but is well understood, it can be saved."

[motion guru]

One additional aspect of PC Resident control boards that causes me to cringe . . . how many wires are connected to that board?

Unless it is something like Sercos or Delta Tau's MACRO where all you have to worry about connecting to the PC is two fiber optic cables and nothing else - you are courting disaster in a factory setting.

Ethernet is fine - USB is fine - but dont have a system where you have encoder wires, amplifier interface wires, limit switches, I/O going to a PC chassis . . . if one shorted out 240V or 480V servo drive blasting the PC wont make a believer out of you then watching a disgruntled maintenance tech swapping motion boards in a PC will.

[RalcoBe]

I agree about the advantages of an Eternet communication over ISA/PCI/AGP/...
but still find you can't compare both.

In the first case you build a network of 2 machines, in the second case you have 1 machine. A computer extended with a board, or a controller board extended with an HMI.

Choosing for the first option has its disadvantages too. Synchronisation, timing, multi-developer issues, and who says ethernet still exists in 10 years?
Of course there will be ethernet-newthing converters, but we all know the problems with rs232-usb converters.

[CarbideBob]

One little problem with Ethernet or USB is that neither one is deterministic "hard" real time. Which lead to systems that work correctly 99.99% of the time. It's the rare glitch that gets ya and they can be very hard to find.

Would you want to fly on an airplane or drive a car controlled by either protocol?

Of course if you are using off the shelf MS windows operating systems you have already lost this battle.

Yes, there are ways around it but too many people don't spend the time and money needed to add safetys for problems that occur once a year.
Bob

[motion guru]

Again, I would never use a PC for anything other than HMI and connectivity - the control should still reside in a dedicated controller rather than a real time kernal running on the PC CPU.

I just did a laser table / controller for a bicycle components manufacturer - completely standalone controller with a simple network connection to the Host PC - they can pick Ethernet, USB or serial cable to connect to the stand alone controller - if the PC takes a dump while the machine is running, it recognizes that the Windows machine is off picking daiseys and it brings the system to a controlled stop and waits patiently for a reboot - no damage parts, and it can pick up where the PC left off without a hitch.

The controllers coming out today have built in connectivity options out the wazzu - even embedded web servers that you can load with HTML and Java scripts to launch a built in HMI from any internet browser. We are building a machine now that might be kind of amusing in that we can have it send updates on it's production performance via. twitter.

[CarbideBob]

Originally Posted by motion guru

Again, I would never use a PC for anything other than HMI and connectivity - the control should still reside in a dedicated controller rather than a real time kernal running on the PC CPU.

Amen to this.
Can there ever be a "real time kernal" on a PC?
The x86 and PC architecture pretty much makes this impossible.

Years ago programmers paid attention to clock cycles and latency.
Nowadays you have to beat this into peoples heads. Too many people trained in software with no idea of how the hardware works. If the cpu is "busy" it won't matter what the next instruction is.

But the problem is that this level of analysis is expensive and the marketing department wants to ship product now, or in the case of a low cost hobbyist wood router an occasional "failure" is considered acceptable. Way different world when you are moving 2-10 tons of steel that will kill someone.

Bob

[BadBeta]

As the original theme of this thread was NI Labview, and not specifically the Windows based PC version, I'll just briefly reiterate that Labview has proper real-time alternatives. (And I do believe they have some 10 year part support guarantee, but I'll not state that for certain).

That said, I have an ambigious feeling about NI and Labview. I think my general describtion would be along the lines of "visionary, passionate, technically clever with huge potential" matched with "greedy, arrogant, non-standard and somewhat of a technical chop suey of products". To me their pricing policies clearly shows that they are not aiming for the control mass market - then again, you do have their involvement with Lego.

[May]

we don't use NI LabVIEW frequency.maybe i will try to do in this way.

[RalcoBe]

Originally Posted by Winfield

...Run FEA Simulations...
... Must be unnerving as a programmer or expert specialist to realize that a common engineer can accomplish the same set of tasks without knowing every detail of a specialty.

And then, we all look at a great FEA simulation, beautiful presentation of person A.
.
B: "Where did you set your boundaries? witch assumptions did you make?"
A: "don't know, just put the picture in the simulation and clicked start"
B: "I can run a simulation, where the results are different with a factor 10. Reality proves it, the part DID break on another location"
A: "So how would you? I want to see that"
B: "Click, click, here it is"
A: "This program is stupid, and not easy to work with, I don't like it"
...


Don't be a specialist. Try to be.

.

[Winfield]

Well now, the result of last week's competition when we asked you to find a derogatory term for the Belgians. Well, the response was enormous and we took quite a long time sorting out the winners. There were some very clever entries. Mrs Hatred of Leicester Said 'let's not call them anything, let's just ignore them' ... (applause starts vigorously, but he holds his hands up for silence) ... and a Mr St John of Huntingdon said he couldn't think of anything more derogatory than Belgians. (cheers and applause; a girl in showgirl costume comes on and holds up placards through next bit) But in the end we settled on three choices: number three ... the Sprouts (placard 'The Sprouts'), sent in by Mrs Vicious of Hastings... very nice ; number two..... the Phlegms (placard) ..