software for robot controling

[FredCNC]

Hi all. I´m doing retrofit in a old ASEA 6 axis robot,and I pratically changed all servo drives/ eletronics for plain DC motor with encoders and for drivers I´m planning on using UHU step/direction servo driver with PID for low-cost eletronics. But I´m stuck with the software. Do anyone knows a software that would do a reasonably good job?

[unterhaus]

EMC http://www.linuxcnc.org used to have a kinematics module that worked for PUMA robots -- I'm sure the module is still around, probably needs some tweeking. If you don't have very many axes, you can use Mach by pre-computing the kinematics. I know some people have used OROCOS -- open robotics something control software. http://www.orocos.org/applications.html

[FredCNC]

I know all three programs, and EMC seems to be the most promissing. Yet, getting inside and reprogramig a new kinematics is another history. I dont know if it is easy or not, because I couldn´t even find where to do. I´ve been looking around for documentation about how to do it, but nothing yet!
Good tip on the mach!
I will double check on the orocos, but last time I saw, it was still in development stage. At least that was my impression
Thanks for the help
Any more ideas?

Fred

[keithorr]

It's a small inexpensive program to model robots and automation. You can use a robot from the library, or build one from scratch. I didn't bother with the skin, just used the links in a bare bones way to simulate the dimensions of my robot. The documentation is sketchy, but you can simulate movement between waypoints and save the axis joint angles. A movement can be a single point to point move, or a series of waypoints, so your text file could just be one line of six joint angles to move 5 feet, or 250 lines to move 5 inches. The trajectory is a text file that can be annotated and read as Gcode by Mach or others. I am using camsoft with a seven axis galil card. I can't use the program to jog the arm which is the only function I lost when I did the retrofit. The original firmware has the kinematics embedded so you can jog with the end effector maintaining a relative position throughout the movement. Anyway.... the wmv shows a simple point to point move that can be saved and played by the robot.

[GunnarK]

I would be interested in what program you mean in your post. Also, what happens if you have more axis then degrees of freedom? I work regularly with a machine which has eight axis, so for one position of the end efector, there are many ways to reach there. Does the software handle these ambiguities, or is it possible to "pre-program" certain axes ? In that situation, you would e.g. pre-program two axis, so you have six degrees of freedom again for your IK (inverse kinematics).
I am talking about inverse kinematics here since forward kinematics is not different to a machine with a smaller number of axis.

[keithorr]

http://www.kinematics.com/products/e...ist/index.html

I understand. Two axis for x,y base under 6 degree of freedom robot arm.

You would have to build a model of one to see how it works. There are values for axis priorities, but I have not used the feature. There is no limit to the number of axis you can have.
If you have high speed internet, two videos from my shop;

http://www.keithorrblowpipes.com/ima...inishing2a.wmv
http://www.keithorrblowpipes.com/images/Norton.wmv

[teebee]

Hi
I've done a small robot arm for my thesis a few years ago, including a control program in VB (the arm was driven by model servos). It can compute FK and IK and also does PointToPoint moves and Linear interpolation (never had time to ad circles). I still have the code so if U are interested send me a mail...

One more think that should be said... Before even thinking of using parts of the code U first have to learn at least basics of slovak language, since all the comments in the VB code are in slovak...

[trubleshtr]

So is this software (robot assist) just for simulating, or can it actually run a robot?
I am a little confused.........
I have 2, 4 axis, robots that need new control/software.

[GunnarK]

I had a look at www.kinematics. com, thanks for the link. Unfortunately the Demo-Download page didn't work, but I was told they are working on it. From their webpage it looks like they also support setups > 6 axis, which could be interesting to try out.
@ Keith, I also saw your two movies about the glass cutting and finishing. Very interesting indeed. did you buid the grasping mechanism yourself ? Looks very nice, expecially concidering that glass might be more stable than an egg, but certainly not unbreakable... Do you use preassure sensitive elements for feedback, or how do you avoid breaking the delicate material?

@ teebee: How many servos did your arm use ? If it is less than seven, the kinematics is _relatively_ easy (I wouldn't say trivial :-), since for every position (excluding gimbal lock) there is only one set of servo positions how you can reach there (different story, if you want to reach directly above the first center of rotation, for example, since moving that axis doesn't necessarily affect the position of the hand in a way which does matter)
If it has seven or more (independant) axis, then you have more than one set of servo positions to reach a certain pose, but one might be more suited than the other, e.g. when continuing a move from one position to the next.

@trublshtr: From what I saw on the web page, the test versin is more for playing aroung with and trying out their kinematics engine. But since they have the engine, they might be willing to sell that as a product. But beware, that might be pricey. since making a workable kinematice for arbitrary robot configurations is non-trivial...
It might be an idea to look into EMC for that, since i heard (but didn't look into that myself just yet), that EMC would support some konfigurations for inverse kinematics as well. That might also be a way to go for me, since I will be starting with less that six axis anyway...

P.S.: If you want to have a look at which machines I use to work with, have a look at http://www.mrmoco.com especially the Cyclops, that's the machine I learned on...
Can be quite scary to have that running past yourself at 3 meterrs per sec, let alone the rush of wind...

[teebee]

Originally Posted by GunnarK @ teebee: How many servos did your arm use ? If it is less than seven, the kinematics is _relatively_ easy (I wouldn't say trivial :-), since for every position (excluding gimbal lock) there is only one set of servo positions how you can reach there (different story, if you want to reach directly above the first center of rotation, for example, since moving that axis doesn't necessarily affect the position of the hand in a way which does matter) If it has seven or more (independant) axis, then you have more than one set of servo positions to reach a certain pose, but one might be more suited than the other, e.g. when continuing a move from one position to the next.

It was a 6-axis robot (see pics below).
You are a bit wrong about IK solutions... In fact a 6DOF robot arm's IK is not "relatively easy" at all... For each point in the robot's workspace there is more than one possible joint configuration. "NORMAL" points can be reached through up to 16 different configurations (theoretically - for a robot w/ unlimited axes movements possible) - in reality 2-4 of this joint configurations can be reached. The good message here is that these configurations are quite far away from each other (at least most of the times)
Some special points called singularities can be reached by an infinite number of configurations (for example directly above the first axis of rotation or if you strech the wrist so that axes 4 and 6 are colinear).

In both cases some sort of optimalization has to be used to find the one configuration which will ensure smooth movement of the arm. Often this is based on minimizing the overall axis travel to reach this point. Since for normal points the configurations are far away from each other, you just need to simply drive the arm to the neerest one of them.

Solving IK for more than 6 DOF is possible, yet it requires an iterative aproach (gradient-based or optimum-based methods or something like that).

For 6 DOF IK can be computed explicitly which means faster and more precise than iterative. Unfortunetly these explicit equations have to be writen for each robot arm configuration separatly.

[GunnarK]

teebee,

thanks for pointing this out, and clarifying my above post. I was i bit optimistic above, you're right. My concern was rather about the singularities, I must admit. With a 6-dof, you have only a certain amount of singularities, as you describe. With > 6 DOF, basically most of your workspace becomes a siongularity. It never fails to amaze people, when basically the whole rig moves, and the camera (my "tool") stays in the same place. This is very nice for programming purposes, since the pictures in the camera stays the same, which is wat we want, and still I can reorient the axis positions, so that I might be able to reach the next pose more easily.
But certainly you are right about the explicit aproach, which is easier, but still a lot of work.

Looks like a very nice arm, did you construct it yourself ?

[teebee]

It was my master thesis... did work quite well although the model servos are to inaccurate (only a resolution of about 0.7deg) and so the precision at the outer perimeter was onlly smthg like 3-4mm... VERY INACCURATE for a robot

the kinematics however worked well and in linear motion you could see that the robot realy tries to follow a straight line, just the actuators didn't allow it... :-(
I wanted to build an updated version with servos, but unfortunetly the budget at my uni didn't allow such fun... will build it probeably just for private fun when I found time and funds...