I am really new to the topic. There doesn't seem to be lot of overlap between Industrial robotics and Hobby robotics (atleast in certain areas like control etc).Please correct me if i am wrong. I actually tried going through Fanuc website, and most of the content is restricted. I would like to know if there is any course on how to operate Industrial robots? its PLC programming? or any application specific course etc?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
7
To directly answer your question, you can take a look at some of the ROS MoveIt! Videos. Check out the Udacity robotics courses.
To give a more opinionated answer...
I've worked both with industrial robots and mobile robots, so my opinion is that if you learn mobile robotics concepts including manipulators and become familiar with robot system design, you'll be able to do well in industrial robotics. However, industrial robots are typically programmed in a very different way.
Mobile robots are designed to sense and respond to their environments.
Industrial robots are primarily designed to follow pre-planned instructions and manipulate their environments.
So industrial robots are typically not designed with a wide array of sensors and complex conditional logic. It's a lot of sequential programming in specialized languages such has RAPID (ABB), KUKA Robot Language, KAREL, Fanuc TP (each robot manufacturer has its own, and there are some industry standards). It's a lot more like CNC programming, but without G and M code and rather a list of human-readable sequential steps. You can watch YouTube to get up to speed on any of these languages, but none of them are that revolutionary. If you're already a programmer, you'll probably cringe a little and roll your eyes when you see these languages.
The skills in mobile robotics transfer over, and you can even make big improvements to the industrial robotics world by bringing in the more traditional skillset (which sensors, algorithms to use, metrology, etc.)
Most training on industrial robots is either done at universities that have partnerships with the robotics companies or at companies that have purchased these robots. Thus it's hard to get into the details of this field without being there. But you're not missing much in my humble opinion. I've been through those trainings and I mostly rolled my eyes because industrial robotics seems to be miles behind the rest of the robotics industry in terms of how the user would program them. Thus just learning robotics in general will help a lot.
Most industrial robots just perform DSP on the motor signals to figure out where they are. It's amazing and kind of scary, but there's no physical positioning feedback. The error adds up through the chain out to the end effector and it's a pain. Some of the more advanced robotics groups thus measure the end effector from the outside to correct these estimates, and that's done with expensive software rather than someone actually programming anything.
In general, if you learn MoveIt!, you've read about robotic arm path planning, and you are familiar with robotics concepts in general, you'll be ready to jump into the world of industrial robots, and you might even cringe a little when you see how the robots that run factories really work.
-
$\begingroup$ I mostly agree. I do not agre with ros mentioned as an example for learning to deal with industrail robots. Also industrial robots do not perform DSP on motor signals to figure out where they are. They have encoders in all joints and can achieve 10 micron positioning precision. The reason they do not have an external camera system is: they do not need one. There is no "traditional skillset" that is missing in industrial robotics as you mentions. Moreover, industrial robot companies enter the mobile robotics markets, and not the other way around $\endgroup$– 50k4Apr 21, 2017 at 13:49
-
$\begingroup$ Furthermore do not underestimate industrial robotics. They are behind research, this is true for all fields, not just robotics. However, they have advanced capablities needed in the industry. E.g. countour finding for welding applications, builtin state machines for easy task dispatching from externally connectes cell controllers, cooperation functions where one robot works in the moving reference frame os a different robot. $\endgroup$– 50k4Apr 21, 2017 at 13:52
-
$\begingroup$ The KUKA robots I worked with had no encoders- 100% of the pose data came from DSP, according to the KUKA reps I worked with. Same w/ ABB. The traditional robotics skillset includes programming a robot to perform Bayesian state estimation, sensing, and decision making. Industrial robots in my experience almost never use these skills, but rather are programmed purely imperatively. Do x, do y, do z, go here, come back. Thus path planning knowledge is somewhat relevant, but it's almost always done offline, and almost always done by prebuilt software GUIs, not by a programmer writing code. $\endgroup$ Apr 21, 2017 at 22:17
-
$\begingroup$ I would say that because industrial robotics is "good enough" with this offline-planned method and factories can get good enough consistency that these pre-planned paths are highly reliable even with no sensors at all, it stifles innovation in the field. Industrial robots don't take advantage of all the great robotics technology in the rest of the industry that's a bit more cutting edge. Responding dynamically to sensor data isn't "research," it's a tenant of robotics, and one only very new industrial robots are beginning to take advantage of, outside of safety applications. $\endgroup$ Apr 21, 2017 at 22:21
-
$\begingroup$ And I totally agree with you that there are a few really great emerging technologies in industrial robotics and it's starting to catch up, but the majority of industrial robots I've interacted with aren't so glorious, and those that are... are very buggy! $\endgroup$ Apr 21, 2017 at 22:27