Delta robot extruder / sorter - building from scratch

Hi to all !
This is my first post to this forum.

We ( four of us ) are planning to build and implement pick and place delta robot that would be similar to ABB'a flexPicker.
Of course, not so complex and with lower pickup rate.
I hope a can get some good technical guidelines on this forum so i will write my questions in full detail :-)

Stuff that i know:
Conveyor track size : 50 cm
Conveyor speed : adjustable
Available space between conveyor track and roof of the sorting container : 120 cm.
Material for sorting: colored plastic lid from plastic containers ( green and blue should be extracted )
Material size: 10 cm in diameter
Weight of lid : under 70 gram's
Wished pickup amount: 40 - 45 PCS per minute
Drop place: right and left from the conveyor track.
12V Vacuum pump
Time to develop this project: 6 - 8 months

Stuff that i think i know :-) :
The idea was to combine arduino and raspberry on the way that we would use RSP for vision system and arduino for IO stuff + RADDS controller.
RSP sends coordinates and pickup time, and arduino does the rest.
I was thinking on using aluminum parts for arms and steel / plastic ball joints but if possible carbon fiber will be used instead.
Arm dimension ( not really calculated yet ):
--> biceps --> 30 cm
--> forearm --> 70 cm

What i dont know
- how strong stepper motors should i use in this setup ?
- as far as i know RADDS supports up to 15A so i guess i can use this controller without problem ?
- from mechanical point of view, can i use this setup to archive this pcs/second ratio ?
- do i have to use gearbox to obtain this speed ?

Software part
I was thinking to hookup some wide angle high resolution USB camera, at the begginig of conveyor track.
Object would be read by sw ( sw would store coordinates), and speed sensor would be used to define pickup time.
So, x= 10 cm , z=20 cm, pickup time = 0.5 sec.
Is this good approach ?

I could write bunch of additional data and ideas, but let's start from somewhere :-)
Thank you all for your future suggestions and support !

Arnix

Which part of a Delta is the Bizeps and Forearm? Did you mean SCARA kinematics instead?

IIRC there was a succesful kickstarter campaign about a color recognition system based on a microcontroller ( RPi? )
Someone else please chime in here for more details/links?

The steppers commonly used in RepRap world will easily lift a part of ~70grams, but I'm concerned about the vaccum part.
Don't you need a surface scanner to find a good landing area for the suction cup? Bigger cups provide more lift capabilities, but require a good landing area, too.
Can you guarantee, that the suction cup will find a spot on the part, where it can connect vertically? Or do you need a few more axis to aim the tool head in the right direction? Then an 8bit Arduino will probably be the wrong choice. FWIW Duet boards support X,Y,Z,U,V,W kinematics.

Last but not least: why use a suction cup instead of compressed air? Use a conveyor belt made of metal chains and blow air through it from underneath. The air valves aim to the right or left, to blow the part in the right direction.

Quote
o_lampe
... FWIW Duet boards support X,Y,Z,U,V,W kinematics...

And also A, B, C axes on the SAM4E-based ones (9 axes maximum).

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Hi guys !
Am terrible sorry for late reply. I catch some cold and i was like zombie for few days:-)

The idea is to build standard 3 armed parallel delta robot.
Forearm would be the first ( smaller arm, connected directly to motor ) and bizeps is actually connecting rod...

Vacuum pump:
- to be honest i didnt think much about that but now when you point it, i think i should create some different suction cup with some
light material that could "fit" better to surface. Conveyor track is already set and it's standard rubber one, so i can not change this part.
If i would blow compressed air onto objects, i would have to remove existing setup and this is unfortunately impossible.
I was watching some y.t. video where flexpicker was used for picking pancake. There was also a vacuum pump but it looks like that the
"object" is slightly pressed to ensure correct pickup...

HW part:
- if possible i would avoid additional motors but if necessary we can implement that to.
- at the moment i would use just basic model so i can check how much speed and accuracy can i get.
- based on that, there is question about motor controllers and motor power.

O_lampe mentioned that 70grams would not be some big problem.
Can you share a bit more info on that ?

What kind of motor should we take into considerations, do i need gearbox to get bigger speeds and finally, RADDS or Duet boards ?
I was searcing for RADDS and it looks like it's well covered and tested, but i can not find how strong motors can be used with this controller....

I can not wait to start with this project but our recurses are limited so i have to make some bom list before we start :-)
Am really glad i found this forum because it looks like this is the right place for every possible question about delta robots :-).

Arnix

Edited 1 time(s). Last edit at 05/01/2018 05:37PM by Arnix.

I'd go for DuetWiFi, simply because it has WiFi onboard. And their open source firmware development is alive. The RADDS guy can't help you much I'm afraid.

Speed and motor torque are two opposite goals. You can't have both. Acceleration is a third parameter you have to deal with.
With a bit of ballistic experimenting, you could find a clever way to throw the picked part in the right direction. That way the tool head has to move only half the way. Saving time to pick up the next part.

To help you choose suitable motors and supply voltage, you can use the EMF Calculator at [reprapfirmware.org]. The general principles for achieving high speeds are:

- Use low inductance motors (but make sure your stepper motor drivers can provide enough current to them);
- Use at least 24V supply voltage;
- Don't use 0.9deg motors unless you really need to for positioning accuracy. 1.8deg motors can do higher speeds for the ame supply voltage.

Edited 1 time(s). Last edit at 05/02/2018 03:21AM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Guys, what can i say... thank you for suggestions and ideas !
I will go then with duet board.

@dc42:
something like this:
[F0AAOSwDNdV6T-n" target="_blank" rel="nofollow">www.ebay.com]
I will check the calculator too...

@o_lampe
The throwing idea is great and with some head alignment it could work very well. Thnx !

Arnix

That looks the same as this one [www.omc-stepperonline.com] which is what I use for testing Duets at maximum current. It's a good choice to drive from a Duet if you need a Nema 23 motor, however I suspect that a Nema 17 motor would be powerful enough for your application.

Edited 2 time(s). Last edit at 05/02/2018 12:25PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Hi dc42.

I dont have to use nema 23 motor.
The "only" thing that matters in this setup is speed. Am aiming to get between 40 - 45 pickups per minute ( 50-60 would be my final goal in near future ).
If this can be done with nema 17 motor and DUET board, then we can use this. I was thinking to use 23 motor just in case if we will need to add gearbox to increase the speed (maybe am on wrong track with that :-)).

Other idea is to remove raspberry and add some linux laptop to do the vision part job.
In that case i can process selected objects faster and arduino can execute his part faster...
This is just my opinion...

Gonna take a break...
Thank you for your time.

Hear you tomorrow !

I found an old link to the color tracking device. It's the Pixy CMUcam5, which is widely available already. You have to watch their promo video, it's fun.

It comes with SPI, I2C, digital and analog I/O ports. Easy to integrate into RRF, David?

(OT: Just thought, I could build a lawn mower bot with the Pixy. Haven't seen any similar videos on Y.T. but some robot arms sorting machines!)

Edited 3 time(s). Last edit at 05/04/2018 05:55AM by o_lampe.

Quote
o_lampe
I found an old link to the color tracking device. It's the Pixy CMUcam5, which is widely available already. You have to watch their promo video, it's fun.

It comes with SPI, I2C, digital and analog I/O ports. Easy to integrate into RRF, David?

That device looks interesting, assuming it can really do what the promotional material says. It seems to be available from several distributors, including AdaFruit and SparkFun. The easiest way to interface it would probably be via SPI using the daughter board connector or the expansion connector, if the wires to the device can be kept reasonably short.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

There's another device called OpenMV, programmable in Python.

I have to finish the BOM list so i will try to summarize all your suggestions :-)
But before all, let me just say that speed of the robot is the main thing so everything depends on it ( servos, vision, gearbox.... )

1.) Pixy
- I saw this project before some time but i give up because i was thinking that there is no way that arduino can process pixy data and control the stepper motor controller, without loosing processor time and memory. Maybe,am wrong... I dont know what would be fastest recognition --> run servo, respond time.

2.) OpenMV
- didnt heard about that but it looks is a bit stronger then pixy.

3.) On what distance can such camera detect some colored object ?.
In my case i have to put camera as high as possible, so i can get bigger working area...

4.) Would this vision system detect let's say, white color ?

5.) Should i order nema 23 or nema 17 servos + duet board ?

Conclusion:
If i understand your suggestions correctly, DUET board should be used in any situation, nevertheless of vision part.
If you both think i can avoid some PC based solution i would gladly use something that is already available...

Anyhow :-), all comes down to create fast robot :-).



Arnix

Edited 1 time(s). Last edit at 05/04/2018 05:43PM by Arnix.

The data stream coming from the MV-cameras is not that much and in your case the objects only move in one direction and with constant speed.
Whenever the tool head is "ready" you pick the coords of the next best part, add a precalculated offset and move the toolhead there.
The parts probably have different heights, so the toolhead has to follow the part until its picked. You have to find a way to determine, whether the suction cup is in contact or not. ( maybe a shunt resistor in the dc-pump powerline? )

The cameras can detect white parts ( not transparent ). You have to find a good spot for the camera. Vision must not be interupted by the delta arms or toolhead. BTW: The toolhead can have a colormark as well. That way, you can have closed loop control for "the wedding " of suction cup and part.
The openMV has several lenses to choose from, but it's more dedicated to face recognition and following pupils it seems.

I'd start with NEMA23, because you can't go wrong with them. Design the Delta frame in a way to be able to use belt pulleys of different diameter. That way you can modify speed vs. torque without replacing steppers.

edit: I just read the wiki of Pixy and they can't detect white.

Edited 2 time(s). Last edit at 05/05/2018 11:01AM by o_lampe.

Hi !

0.)
Design the Delta frame in a way to be able to use belt pulleys of different diameter. That way you can modify speed vs. torque without replacing steppers.
** i agree with that :-)

1.)
- what kind of nema23 stepper motor would you suggest to use, with duet board ?
- i saw that DC42 has done some tests with this board, so if you read this, please share your suggestions too.
- also, dc42 has suggested one stepper motor but i dont know if i can use this with duet board ( dc42 --> comments ? ).
- i didnt work with duet board, can i hook up some 24V nema motor to this board ?

2.)
Pixy / openMV
----------------------
a.) Can those sensors used in pixy or openMV detect let's say light blue, or light green or any "light" color ?. I mean, how much color in some sample, is enough for positive detection ?
b.) What do you think, would pixy or openMV be OK for detecting color samples 80 cm away from delta robot base ?


Arnix.

You can use the Duet with 24V power. The maximum motor current that the Duet can supply is 2.4A, but as it's usual to run motors below their rated current (otherwise they get very hot) it's suitable for 2.8A Nema 23 motors, such as the motor I suggested.

You can use the calculator at reprapfirmware.org to work out the maximum speed you will be able to get without loss of torque.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Sounds like you're building a Delta-based pick and place machine, correct? If so, the old FirePick project might be helpful.

I can only refer about the Pixy. It can detect light colors, especially in a closed environment like yours, where YOU pick the source of light. The camera can be tuned to get the most out of it. When you teach new color signatures, you hold the part in 5-30" distance. Once it has learned the color, an area of 4x1 pixel is enough to detect it.

In my CNC mill, I use 23HS8630B steppers with 3A max current. But they run very well with lower current. These are 6-wire steppers, so you have the choice to wire them in high torque/low speed mode or vice versa.

Hi guys !
After some thinking and consolidation with my colleges, we decided that we will go with openCV but we will use pixy as backup.
OpenCV is great peace of sw and it has bunch of options so i think that this could help us.

Motor:
- ok i will take a look

Duet board
- @dc42 --> are you selling duet boards ?.

@cobalt griffon
- it's some sort of pnp machine. It looks like am not the only one on that rsp - arduino track :-)

Arnix

Edited 1 time(s). Last edit at 05/08/2018 07:46AM by Arnix.

Quote
Arnix
Duet board
- @dc42 --> are you selling duet boards ?.

See [www.duet3d.com].

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

DUET3D quote:

If your motors are rated above about 2.8A and you are using the Duet WiFi, or above about 2A and you are using the Duet 0.6 or 0.8.5, or if they need higher voltage than the Duet can provide, then you need external stepper motor drivers. These generally have optically isolated step/dir/enable inputs. For example, stepper motor drivers rated at up to 5A using the TB6600 stepper driver chip are widely available on eBay.
If the drivers require no more than about 2mA @ 3V on the step, dir and enable inputs, then you can drive them directly from the expansion connector of the Duet.
******
So, if i use 23HS8630B as test motor, do i need to add TB6600 ?
If so, what would be suggested choice ?

Shortly:
I just want to see how will this, or any similar motor work at some maximum, but still safe, level...

Arnix

That motor is rated at 3.0A, so if you drive it direct from the Duet at the maximum 2.4A current then you would still be driving it at 80% of its rated current. However, I really don't think you need such powerful motors for the device you are building. Also the one I suggested earlier costs much less.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

HI guys !

I will get my duet board in day or two and am still waiting for motors.
I took nema23 just because i want to see how strong they are.
OK. now the interesting part :-) Starting with DUET3D.


1.) At the first view i can hook up duet with USB or ethernet so i guess i can send data coordinates on the same way.
But, how can duet process the incoming data ?. Do i need arduino for this or can i just send PC data directly to Duet?
We are writing sw that will be used for object tracking and it will send ( per ethernet or USB ) data to duet board.
But i dont get it how will Duet receive / process this data.

Shortly, object on the conveyor track will be almost always on the different place and in the different time.
E.g. if i send 5 coordinates in let's say 5 seconds time, how can duet accept and process this data ?
Do i need to save it somewhere on the duet board ?

2.) Ok, i have like 100000 questions but i will try to calm down :-)
I was thinking to activate the air nozzle with arduino and on the way that it triggers itself when robot head is under some predefined
level. E.g. some sensor will be hooked up near conveyor track and if the sensor on the head is under this, activate air nozzle...
Just an idea ...Btw. can this be done without arduino, just directly from Duet board ? If so, how ?

Arnix.

Edited 4 time(s). Last edit at 05/21/2018 08:51AM by Arnix.

1. You can send GCodes from the PC to the Duet via USB or via HTTP. If you want to send via http other than through the usual web interface, then an easy way to do this is to use curl. If you want to use the same sequences of commands many times over, you can store those sequences as macros on the SD card.

2. You could use a spare fan output from the Duet to control the air valve, or if it is a high-current valve then a spare heater output + flyback diode. Then you will be able to send a M42 GCode command to the Duet toi turn the air on or off.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

You'd have to split each pickup move in three sections.
1. move to position above part
2. activate air nozzle
3. move toolhead down until contact

The Duet controller has a generous command queue and can take 5 coords in 5 seconds easily, but I'm not sure if the mechanic is fast enough to process these moves in 5 sec. ( worst case is +50cm travel move +50cm slower motion with the part picked up )

I got my duet board !
Now, this is some serious peace of hardware :-)

@dc42
1.) OK, but how do i make handshake with firmware ?. Because am sending those GCodes from our own software, not from DUET web interface.
USB host , Ethernet packages with gcode encapsulation ? Basically,what do i have to do before i can send GCodes to DUET board :-).

2.) Great !

@O_lampe
1.) check :-)
2.) check :-)
3.) problem :-) ----> how can i determine the distance between object and toolhead.
Or better, what would be best/practical way to measure that...

Quote
Arnix
I got my duet board !
Now, this is some serious peace of hardware :-)

@dc42
1.) OK, but how do i make handshake with firmware ?. Because am sending those GCodes from our own software, not from DUET web interface.
USB host , Ethernet packages with gcode encapsulation ? Basically,what do i have to do before i can send GCodes to DUET board :-).

The easiest way to get started will to send the GCodes over USB. The Duet will look like a COM port to your PC so you can send GCodes to it from a terminal emulator, or from a 3D printer host program such as Pronterface.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

@dc42
I log myself on duet forum and i noticed that you are hw designer and sw developer of the board.
Quite nice project !

USB:
- i have to check the firmware before i can continue this part of conversation but what i meant regarding handshake is:
--> before i convert string to gcode, does this have to be sorted somehow.

Do i have to send x,y,z in this or some other order ?
On the other side, how to send timestamp...

OK, I will look at the firmware and then i will ask some meaningful question:-)

Arnix

Edited 1 time(s). Last edit at 05/22/2018 03:35PM by Arnix.

Quote

3.) problem :-) ----> how can i determine the distance between object and toolhead.
Or better, what would be best/practical way to measure that...

You have to install some sort of z-endstop on your toolhead. Then send G1 S1 Zxxxx . That S1 paprameter lowers the z-axis until the endstop triggers.
The endstop could be a flow sensor in the air-tube, but then a missed "kiss" of the suction cup would be disastrous.
Also you have to remember, the belt is still running, so you'd have to move the toolhead in Y-direction too.

IMHO you don't need to convert string to gcode as long as /n marks the end of line. But dc42 surely knows better.

The G-, M- or T- command must be at the start of the line (unless you are sending line numbers too) but the parameters can be in any order.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].