Anyone familiar with the ROBOTICS INDUSTRY? 3

[UncleJess]

Hello Everyone,

Does anyone have any experience working with robotics?

I am a mechanical engineer in my early twenties. My career goal (fantasy...) is to work with robotics one day. In college, I loved all of that inverse kinematics and control jazz. I really enjoy all of the involved mathematical analysis and whatnot.

From what I have seen during my first year working in the real world (I am a recent grad. currently designing custom machinery at a machine shop), aside from spaceships and airplanes, it doesn't seem like anyone gets to do long mathematical analyses. Seems like everything is "shoot from the hip".

I hope that this is not the case in the robotics industry? Can anyone who is working or has worked in the robotics industry educate me a bit? Do ME's in the robotics industry get to use any of the fun math behind robotics or do computers handle all of that now? If computers do handle all the analysis, what else is left for ME's to do? Material selection? I guess my main question is, as an ME, what's out there for me to do in the robotics industry?

For all I know, working with robotics may not be my goal after all...

Thank you very much if you have read this far!

 

[KENAT]

I would think if nothing else ther'd still be work to do in designing the end effectors. I just helped a colleague with this and it was pretty tricky.

(OK that was mostly the fault of the form constraings from the previous design we were replacing but still a challenge)

 

[johnsmith2]

I know the robotics automatic control is commonly used in automotive or medical device industry.

 

[ctopher]

Computers do a lot of the work now. Check out jobs at JPL or NASA. They are into robotics, they may be what you are looking for.

Chris
Systems Analyst, I.S.
SolidWorks 06 4.1/PDMWorks 06
AutoCAD 06
ctopher's home (updated 06-21-06)

 

[tygerdawg]

What you are looking for is more of the systems-level programming or research areas.

I've got 25+ years working with bots and also directly in the bot industry (industrial bots, not the "battle-bot" variety of homegrown units). I'd say YES those jobs exist, but there are not very many of them. A friend of mine told me that there are only about 600 people on the planet who actually understand transformational mathematics. He was joking, but there is some truth to it.

The math is well-defined, not much new there. Some companies are developing new arms every few years, but the competition is stiff to be part of those teams. And the robot industry is cyclic, and it dumps its engineers on the market through layoff every few years.

The majority of robot users can't handle the math, and the robot industry has catered to the lowest common denominator: creating simplistic teach-pendant based programming languages that are easy to use, but sacrifice a lot of transformational functionality. It's just beyond most people. The biggest use of robots is welding (path & spot) and there's not a lot of sophistication there. Then there's assembly tasks, and again pretty simple. The fun stuff is the swoopy-curvy path applications that require a knowledge of sophisticated math and programming skill to integrate into an overall machine function. But my experience is that when you get hooked, you're really hooked.


TygerDawg

 

[KENAT]

Probably should have said the application I was looking at was Semi, handling wafers.

Nothing too fancey though, in fact they make some of the welding robots I've seen look fancy.

Ken

 

[UncleJess]

First of all, a big thank you to everyone who has responded! I greatly value any information on this topic (In fact, I keep a special binder for this sort business).

tygerdawg,

When you say that I am looking for "systems-level programming," what exactly does that mean? In other words, what does someone in this position do (I am not familiar)? Is this a job for an ME (sounds more like it would be for CS to me)?

I get the impression that most challenges of creating a robotic manipulator have been solved already. Where do you see the new challenges in robotics then? High speed? Accuracy?

A respectable professor (in my eyes) once told me that, "A robot that does not work using the inverse kinematic solution is rubbish." Do you think that is true? "teach-pendent" doesn't fall into that category does it?

So far, I am hearing that welding robots are not very interesting...What are some examples of these fun "swoopy-curvy path" applications? When do you need something like that?

Thank you very much!

 

[GregLocock]

"A respectable professor (in my eyes) once told me that, "A robot that does not work using the inverse kinematic solution is rubbish." "

A repectable engineer once told me that an engineering professor who hasn't worked in industry is likely to have some silly opinions!

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Beggar]

I'm going to have a phone interview in the next couple days with a robotics company.

Wish me luck. If I make it through the forthcoming hoops and actually get a job, I'll give you some feedback from the inside.

--------------------
How much do YOU owe?

http://www.brillig.com/debt_clock/

--------------------

 

[MintJulep]

When I think swoopy-curvy path, I think auto painting.

However, I thought that most of those were programmed by digitizing the motions of expert human painters. No math needed.

 

[Beggar]

By the way, one of the things that they specifically mention in their ad is knowledge of kinematics and experience with mechanism design.

It sounds like they want the engineers to at least have the knowledge to validate the computer output.

--------------------
How much do YOU owe?

http://www.brillig.com/debt_clock/

--------------------

 

[GregLocock]

The other swoopy ones are the ones that apply sealant from a gun (not that I'm unimpressed by welding bots either) and then stick the windscreen in.

Somebody needs to understand the kinematics to stop the joints locking, even for a path follower.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[tygerdawg]

A few of the actual engineering jobs I saw being done at the robot company:
robot arm link design, design of reduction gears for each of the joints, electrical controller design for servo amps, power dist & control, operating system design including math modeling of the arm links + kinematic calculations + path planning + servo commands + human interface design, testing testing testing, servo tuning, customer software customization, manufacturing engineering to get it all built, customer support, sales support.

Swoop-curvy: deburring, routering, laser welding/cutting of a very precise nature on 3D path with orientation changes, waterjet apps, precision painting of gold-laced paint for cell-phone covers, semiconductor apps where part orientation through a pick-place path is critical at all points, precision polishing, dispensing.

I don't think all challenges have been solved, just look at the new models that are introduced yearly. But new development must be fit into the economic situation. One of the next big challenges is to make programming easier, have more functionality without being more complex, offline programming packages, tighter integration of CAD data to create locations, vision applications that will result in self-teaching paths and obstacle avoidance.

Welding robots are not very interesting....to me. Make your own judgement. As far as your professor, don't believe everything you hear. Including things I say here in this forum.

TygerDawg

 

[SylvestreW]

worked for a while at FANUC, on the customer applications side of things. IOW, I helped make sure the system the customer got did the job they wanted.

I can tell you this much, once the system is built, all the "theory" you apply is remembering the cartesian system the robot is in and making sure you don't hit a singularity

Unfortuantely, at the industrial level, you neither see, work with or interact with the mathematics. Typically, you are trying to minimize the # of movements of the path, making sure you don't "jerk" the tool enough to break the clutch and other lowly specialized things like that.

You can get into more advanced robotic modelling but unfortunately it usually means importing a "robot profile" with known operational parameters and just optimizing things.

Now, don't be too discouraged since again, all I have was the industrial experience. I'm sure R&D and/or exotic applications may be different.

 

[SylvestreW]

Actually, I was lucky to get involved in a really fun "swoopy curvy path" project. Bad news first - management screwed the pooch and killed the project few yrs after I left.
Good news - using two robots to polish car rims.

As you know, rims are curved in all 3 dimensions. In addition, the polishing media is used up during polishing. So to get a consistent, repeatable polish on a surface like that is incredibly difficult.
Actually ended up involving an over-head camera to orient the rim (via the air valve hole). A force-feedback unit on the tooling to determine the wear of the polishing media, and a few months of KAREL programming to optimize the system.

The client actually ended up getting a patent on the system, it was *that* good