|
|
Great serie! I looked at the whole week! 
|
|
|
|
|
|
I do feel left out! 
I was hoping programmers were bound to work on robots! :P
|
|
|
|
|
I guess it depends on your interests and where/how you started programming - starting professionally in 'c', having a background in electronics (among other things), I'd likely be more into Arduino, Edison, Raspberry Pi and things I could do with them than someone who started with java - and Ive constantly followed the trends with the MCU's etc being used
I dunno .. 'maybe' 
These days, as pointed out in a recent discussion, you can run Yocto (Linux) on a Pi and then C#/Mono - so its not so hard for anyone to get started, except you have to learn (and I have to re-learn) interfacing through 'primitive' means to analog, digital inputs etc
you're never to old to learn, and all of the MCU's have a getting started package - so wether you start by lighting up an LED or rotating a servo, get into it 
|
|
|
|
|
I don't know what's useful to learn to work on robot... I guess I should look at job add and make a plan from there!!
Well I did wrote my own neuron network in C#!
|
|
|
|
|
Arduino and Raspberry PI are part of my fun time now. I program the Arduino in C from my Raspberry PI and do some Python on the Raspberry PI. The level of devotion of your spare time and toy money are large. Jeremy Blum has some good tutorials on You Tube. A lot of my toy budget has gone into my experiments.
|
|
|
|
|
I think one aspect has been the prohibitive costs of classic industrial robots. Think about all the amazing things programmers could do if we just had the ability to play with these!
When it comes to more android like robots I think one aspect is because 1, people think its super complicated and 2, you need to work with a lot of hardware. You need components and then you need multiple parts to be able to build a more advanced robot, and to be able to do something you need to construct and program all these. Say you just want to build a simple walker you need to invest a lot of time to succeed here, then when you want to do something more you may have to reevaluate your design.
But I think its easier if you take another approach, build more drone like robots with specific purposes.
I for one would love to delve in to drone/robot development but I lack the time and focus at the moment. Got too many pure software projects I wish to do first.
Sometime I plan on learning enough circuitry and electronics to be able to put together my own drones because I got a lot of ideas I wish to test out.
One thing I wish to try is to use sensors/cameras to read the surrounding, input this in a 3d engine to create a virtual representation of the area and basically just use game ai and pathfinding to navigate around.
Sadly there are too many things I lack knowledge wise atm to test my designs even thou I'm sure they would work if done correctly. 
|
|
|
|
|
Member 11683251 wrote: the amazing things programmers could do if we just had the ability to play with these!
I wonder why the idea of programmers playing with industrial robots sends a chill up and down my spine 
«I want to stay as close to the edge as I can without going over. Out on the edge you see all kinds of things you can't see from the center» Kurt Vonnegut.
|
|
|
|
|
Well, WHY optimize? JIT all the things! Open a StreamCoconutPineapple that manages a StreamFreshWater that includes the StreamBufferedFileCharWithLemon that owns the StreamChar to write a command on the RS-422. What does it mean there are strict timings to comply? How old-fashioned, the hardware should be AGILE!
GCS d--- s-/++ a- C++++ U+++ P- L- E-- W++ N++ o+ K- w+++ O? M-- V? PS+ PE- Y+ PGP t++ 5? X R++ tv-- b+ DI+++ D++ G e++>+++ h--- ++>+++ y+++* Weapons extension: ma- k++ F+2 X
If you think 'goto' is evil, try writing an Assembly program without JMP. -- TNCaver
"When you have eliminated the JavaScript, whatever remains must be an empty page." -- Mike Hankey
modified 13-Oct-15 4:41am.
|
|
|
|
|
I am on the same wavelength!! 
Yeah programming a drone movement & vision could be a good start!
|
|
|
|
|
Industrial vision systems for recognizing faults or position of manufactured parts are fairly accurate and some very quick, using something similar maybe with twin cameras and use to input in to a 3d engine.
I recently tried 1234d catch by Autodesk and was surprised by the results, made a few 3d representations of a few of my childrens' toys.
If you can extrapolate 3d information from a vision system or sensors and input in a 3d engine and merge with already existing data you can probably get some interesting results. What interests me a lot would also be the possibility to create temporary special information to compare with already known data to get the system to figure out its location after a restart or similar.
Another step after you got the system up and running would be to have add probability/permanent information and allow the system to start to figure out if things are movable. You would probably have to add machine learning/ neural networking to do this but I can see a system like this have tremendous potential.
|
|
|
|
|
While actual industrial robots have always been costly (and still are - though if you have access to 3-phase power and a big enough garage, there are many used full-sized industrial robot arms being sold for less than the price of a new car, which is really something) - you have always been able to buy trainer arms for much less (generally well under $5k USD).
Think devices like the TeachMover, Rhino Robotics arms, and the Armdroid (and clones).
Still too expensive? Back in the day - even today - it was a bit of coin.
Back then - the cheapest you could do (aside from a complete homebrew - there was one or two of those published in Byte Magazine in the 70s and 80s) was a Tomy/Radio Shack Armatron robot arm conversion. That had to be one of the most hacked on robot toys of the 1980s - besides the venerable Milton Bradley Big Trak!
Today - kids (and adults!) have their choice of low-cost robot arms to play with; most use cheap RC servos for the actuators, and generally have a cost under $500.00 USD or so. Machine vision hardware and software are both low cost (for experimentation at least - if you want industrial quality, expect to pay some big-bucks for the camera(s) - still, the software remains free, using OpenCV).
All in all - today's a great time to jump into robotics as a hobby - hardware and software have never been cheaper - even if you are purchasing everything. If you are going the DIY route, it's even cheaper! There's a plethora of options for your platforms (land, sea, and air!) and controllers/logic, not to mention software and software frameworks: If you want hard-core research level - go with ROS; if you're just a hobbyist, there's NodeBots (NodeJS) as well as GoBots (Golang). If Windows is more your style, there's plenty of frameworks there, too. Or - grab the bull by the horns and roll your own!
|
|
|
|
|
When I attended a Master's level course on Robotics at my institution in India in 1990, they taught us three things - Kinematics, Dynamics and Control of robots. IMHO, those remain pretty much the same, but today with more computational horsepower for those numerical operations.
Path planning, inverse kinematics were some of the mini-projects I had worked on at that time.
|
|
|
|
|
Interesting answer!
Gonna investigate those topics!
Are you working with robots today?
|
|
|
|
|
Super Lloyd wrote: Are you working with robots today?
No, never ever worked on robots after taking that course 
|
|
|
|
|
|
|
Problem is, a program can only program as good as it's written, there is no AI that would be able to learn by other programs to reprogram it self and get better
Although, if we hit that point, we are basically pointless and the war with the machines will begin.
Rules for the FOSW ![ ^]
if(this.signature != "")
{
MessageBox.Show("This is my signature: " + Environment.NewLine + signature);
}
else
{
MessageBox.Show("404-Signature not found");
}
|
|
|
|
|
I've been looking at secondary schools for my daughter this month, when we went round my old school* they had a computer lab for programming robots, in python iirc.
*when I went to the school in the late 80s they had the most advanced computer lab in the county, a room full of Nimbus computers, we used to go in at lunch time and play snake on them.
Some men are born mediocre, some men achieve mediocrity, and some men have mediocrity thrust upon them.
|
|
|
|
|
Super Lloyd wrote: For how long is it going to be the exclusive realm of a select few?
It's not a select few: it's the few that didn't fall for the shiny webpops things. Those who actually studied math because they didn't believe "only basic math is useful to a programmer", those who took some electronic course and know at least basically what a transfer function is, what is a feedback control system and a finite state machine.
That is, many many people who just "learn how to Java" aren't prepared to manage moving parts. Just think of the extreme prejudice against writing pieces of software in Assembler (whatever flavor you need): each person professing that prejudice can't do serious robotics - they can play with Arduino and DIY robots but only at half power.
All the programmers who rely heavily on frameworks because "that is done by the framework" and do not even try to reinvent the wheel are seriously hampered in robotics and embedded systems, as most probably that is one of a kind or one of a very restricted family, so no framework and no wheel does precisely what is needed.
All that I say from my current experience in Industrial Automation and past experience in Robotics.
GCS d--- s-/++ a- C++++ U+++ P- L- E-- W++ N++ o+ K- w+++ O? M-- V? PS+ PE- Y+ PGP t++ 5? X R++ tv-- b+ DI+++ D++ G e++>+++ h--- ++>+++ y+++* Weapons extension: ma- k++ F+2 X
If you think 'goto' is evil, try writing an Assembly program without JMP. -- TNCaver
"When you have eliminated the JavaScript, whatever remains must be an empty page." -- Mike Hankey
modified 13-Oct-15 7:56am.
|
|
|
|
|
This, imho, is exactly the point for today's environment. Maybe not so much in the future...
"Programming" is...well, not easy, exactly, but fairly simple compared to the real world. At the highest level where users see "Hello, World!", a computer is a computer is a computer. Compilers take care of the rest, and hide all those ugly choices like "add r0,r0,r1" vs "add r0,r1" at the assembler level.
Once you start dealing with the outside (of a computer), things get very complicated. The software for a Roomba won't work on a Cyberdyne T-101, and vice versa. Even if the computers were compatible, the task "get rid of that crud" has to be carried out completely differently, mechanically, on each system, which means that low-level commands would have to be different, thus the math, ad nauseum.
It might be different if robotic platforms (and their manufacturers) settle on the equivalent of "frameworks" for both the computational side and the manipulator/motivator side. Kits and microcontrollers like the Arduino, Lego Robotics, etc. are all a move in the right direction, but your industrial automobile welder is still a world away from a home made robotic arm, even though they might perform the same tasks.
Until then, though, it's going to take a lot of low-level work in both software, hardware, and in-between to get even the most basic robotics project up. Some people love this--I'm working on a homemade trirotor drone made out of $5 of sheet metal from Lowe's and pop-rivets--and they will keep it up.
For the average developer, no matter how brilliant, it is usually too much of an investment of time and labor that, in the short run, doesn't really pay off much, even if the pay off is just seeing something move.
vuolsi così colà dove si puote
ciò che si vuole, e più non dimandare
--The answer to Minos and any question of "Why are we doing it this way?"
|
|
|
|
|
Robots are nice and remove a lot of the dirt work as they are already adjusted to move at their max speed given a certain specs on weight into the clamping system but, they are not precise (therefore the offline trajectory/points programming is always something that is half backed) and when you want to program them you need to go deep into the specific software the manufacturer gives the operator/engineer that need to program it.
This means that you have a subset of variables and functions that are somehow limited, depending on the brand you use bad documented and with a bad technical support too.
Anyway, the best thing of programming robots and special machines is to "see" your work moving something real.
Being a roboticist is usually easy if your environment has companies that are working on that field, of course, studying at the university some robotics help to get in touch with companies that are interested in that world. Anyway, there is something you have to be aware of: there are plenty of companies that search for "robot programmers" which at the end of the day means somebody to move the robot at different positions to store trajectories. Coming from the IT world the nice job is the one that implies maths, 3D trigonometry... to make special and nice things with those arms.
When I do need people in my department I usually search for guys/gals who have studied something related to robotics in the University, but even if they have not, I want people with IT knowledge... It's easier to explain what a piston is than what a pointer is... But this is also an important factor when searching for people: I've seen plenty of programmers that have made all their programming job thinking the mechanics are instantaneous and of course this won't work... When programming with sensors and actuators you need to change a little bit the way of working and getting used to that can take time.
|
|
|
|
|
Not quite sure how you mean "not precise" as the robots in the automobile industry seem to be more precise than the humans doing the same job. But with your signature line and you being a roboticist, you know more than I do as I was relying on co-workers information.
Note: I learned to program a CNC machine in college and only did it once at work (with supervision), but have only the basic idea of robotics. I know enough to consider momentum and that is about it.
Mongo: Mongo only pawn... in game of life.
|
|
|
|
|
This is a quite typical misconception on the robotics field: robots have a good repeatability but an awful precision.
After years of working with them I've learnt to expect a 0,3mm precision in all the working envelope as the best case scenario and only in certain robot arms (the smallest ones). And of course after a very expensive recalibration process that can take a couple of days per robot.
You can easily see Absolut/High accuracy robot variants in all the robot brands out there and in those "special" much more precise systems they are offering +-1mm of precision.
when you move a robot to a certain position and store the point you are simply correcting the precision error without even noticing it, but don't try to copy one program from one robot to another one... even in the same conditions the program will be executed in a different position...
A CNC machine is always more precise than a robot (of course if it has been constructed as it should), but the way the motors and the physical conception of a robot is made it is easy to understand the big kind of errors a robot have.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
