Delta Parallel Robot driven by Arduino microcontro

The servos are completely rebuilt. The original control system is replaced my own control loop, implemented in Arduino Duemilanove boards. Also the default potentiometer is replaced by "AS 5043" encoders from "Austria Microsystems".

The trajectory is generated by a Pathplanner, which considers the Dynamics of the Structure. Maximum allowable torques in joints and maximum allowable forces in Rods are never exceeded. (The robot moves at maximum possible speed, constrained by max torque and max forces in rods).

This video explains why this type of path planning is superior:

Delta robots are not that difficult, and can be built by inexpensive components..

I've seen videos of these types of robots used in food manufacture and parts picking/placement (semiconductors and board assembly); your version is pretty cool.

Do you plan on releasing more details (schematics, code, etc)?

:slight_smile:

The company that I work for is an automation integrator and we frequently use delta robots (ABB FlexPickers mostly) for high-speed packaging (usually food) and high-speed material handling (PV cells).

Although they have both reach and payload limitations, under the right conditions they can run at speeds up to 60 cycles/minute. However while it is not uncommon on even the larger ones to only have a work envelope of 800mm diameter x 300mm deep; this is sufficient for many of the biomedical and PV projects that we work on.

When possible we use vacuum cups as the EoAT. Note, adding a forth axis in the form of a rotating wrist can greatly increase the delta's capability.

It looks like there's one Arduino for each servo. Since one already has 6 PWM pins, is there a reason why you used 3 instead of just one?

impressive :slight_smile:

seen some of them them in some backend fabs and they are super fast!

this blows me away...

Reply to "cr0sh": Yes I am planning to release more details. You can subscribe to my Youtube profile, and you will get a notice, when more stuff is available.

Reply to "CSingleton": I also work at a company where we use delta robots (commercial models) - this is what has triggered my interest for this type of robot. I have not added the rotary 4th axis to this prototype, because the control of this axis is trivial. It is mathematically simple, whereas the physical aspects of it is more complex, when you are on a low budget, like I am.

Reply to "Parcanman": You're right. There is one arduino for each axis. I have tried to use 1 arduino for all three axes, and it can be done. However, I like to run the servo control loops at the highest possible frequency, because it improves the regulation (less oscillations, overshoot, etc.). With only 1 Arduino, the processor is pushed to the limit, and consequently, future extensions to the control loop are limited.

Reply to "designer2k2": Once I have calculated the correct dynamical parameters, it will run much faster than this. I will show some videos later.

"incu": ... It also blows me away. :slight_smile: Didn't think the result would be that good when I started.

I would be interested in looking into the scientific literature that you used for path planning. Great work!

I wonder what kind of ball bearing did you use for parallel joints. It was difficult for me to find anything off the shelf that could have worked without limiting the working space significantly.

medved: You're right, it was difficult to find ball joints which don't limit the workspace too much. I have chosen to use the "Associated Ball Joint Male 4-40", which I have modified a bit with a knife. I have cut a little bit off on the plastic cup, so the angle span gets larger. The reduced cup is not a problem, because the parallelograms are kept together with rubber bands.

Ball joints:
http://www3.towerhobbies.com/cgi-bin/wti0001p?&I=LX2546&P=7

There's not much litterature available about path planning. The method I use, I have developed myself, all the way from scratch, and this process has almost killed me :). There are so many aspects which you need to consider if you want to use a dynamic model for you path planning, and the problem is that you don't know which problems you will face when you start. Basically, everything is Newtons 2nd Law, vectors, and trigonometry.. There's another approach however, and that is not to use a dynamic model. The consequence of course is that you will have to downscale the overall path speeds.

Great little robot! Cant wait to see it do some work :slight_smile:

Very nice

I have created a web page with more information about the robot:

juan,
I have a project to sort oysters by weight, but I need to position them
quickly and automatically. Existing commercial delta robots are too complicated and expensive....I think your approach would work great...
how can I get more information from you?

You did a nice job. Your approach appears to be very systematic and thorough. I would appreciate if you are willing to share some of the ideas with other hobbyists. Say for example what kind of 'Pathplanner' environment are you using? What kind of parameters other than torque are used in simulation? What kind of equations are used? Thanks.

I can understand that you would like more information about the mathematics of the Dynamical Model. The problem is that this is more than 1 year of work, and to describe it, would require something like a phd report. Vectors and trigonometry is a good place to start...

Great work Juan. I will be following your thread for any details.
Thanks