SAPE Single Arm Printer Experiment

Getting started on a new single arm project. Firmware is finished and I have an early prototype built. I am using my worm wheel/gear design from my gdr project. I focused on an extremely low vitamin design. Only 2 bearings for movement, bolts, PVC and printed parts.
Short homing test video:
http://youtu.be/ICIe5_Wij5g

A few shots:


Bearing:

Love it. I imagine everything is pretty rigid in the plane of the arm. Quick shoulder accelerations are probably out. You can beef up the arm, go slow, or rotate the table instead.

Edited 1 time(s). Last edit at 06/02/2014 04:34PM by nicholas.seward.

---

ConceptFORGE
Wally, GUS Simpson, LISA Simpson, THOR Simpson, Sextupteron, CoreXZ

Great fork of your previous design.

a better way would be to mount both the motors on the base .. something like this

forums.reprap.org

Very cool!

Am I the only one that see's it evolving into a backhoe Remotely, and pragmatically dig holes in the yard, scale it up and dig a pond, or a basement, switch the backhoe out for the 3d cement extruder to finish the job! Mount a chair to it for a flight simulator The worm gear will be inherently safe if you loose power too.

Too Funny!

I just finished rebuilding my DIY tractor backhoe attachment last week. I never though about putting a seat on the bucket and running the joysticks out to the seat and taking it for a joy ride.

Quote
A2
Very cool!

Am I the only one that see's it evolving into a backhoe Remotely, and pragmatically dig holes in the yard, scale it up and dig a pond, or a basement, switch the backhoe out for the 3d cement extruder to finish the job! Mount a chair to it for a flight simulator The worm gear will be inherently safe if you loose power too.

LOL Epcot did that with a roller coaster simulator and really big kuka arms.
kuka

Quote
see3d
Too Funny!

I just finished rebuilding my DIY tractor backhoe attachment last week. I never though about putting a seat on the bucket and running the joysticks out to the seat and taking it for a joy ride.

Quote
A2
Very cool!

Am I the only one that see's it evolving into a backhoe Remotely, and pragmatically dig holes in the yard, scale it up and dig a pond, or a basement, switch the backhoe out for the 3d cement extruder to finish the job! Mount a chair to it for a flight simulator The worm gear will be inherently safe if you loose power too.

This is a great "thing" that you have created.

Some one clever could modify it to feed a person who is unable to feed them selves There is software that in real time can give feed back of topology of a dynamic environment (the hungry human) to avoid collision. Or place reflective markers near the mouth to identify the target.

Scalable KinectFusion
[www.youtube.com]

LazeeEye
[www.kickstarter.com]

Edited 1 time(s). Last edit at 06/03/2014 02:21AM by A2.

This does look awesome, though I'm guessing the programming to keep it level and on a g-code designated path will be a bit painful.

Also, while it looks awesome and easy to make, your Simpson-style printer looks much more stable.

Really nice! Reprapable robot arms would be cool for a lot of things... The way computer vision catches up atm you could soon build a robot that can work in chaotic environments and use different tools to pick up stuff, tidy up your room, clean windows. Oh and 3D print of course

Quote
Feign
This does look awesome, though I'm guessing the programming to keep it level and on a g-code designated path will be a bit painful.

Also, while it looks awesome and easy to make, your Simpson-style printer looks much more stable.

The inverse kinematics coordinate translation is all done in the controller (ramps 1.4) . After writing the software for my GDR, the SAPE was not much more difficult. The Simpson style printer (grounded delta) is inherently more stable because of its geometry but it has the same singularities and it relies heavily on symmetry. Symmetry is a cold hearted bitch that will rob you of 6 months of your life leaving only emptiness in your soul.

First print, not so bad. Lots of room for improvement


Youtube candy

Really awesome, thanks for the video and please keep us updated!

A robot arm that can print would offer many interesting possibilities, you could have it rotate 360 degree to move to a new print surface after finishing an object, or have it moving around. Then if you could scale up the design, you could have it print walls or houses!

Is it possible that the Bowden tube is pushing the arm off location?

To increase stability, could you make two opposing arms work together?

I'm looking to build something similar, first on a small robot arm, then build a very large one or use an existing robot arm. Can you please share your firmware? Or share how to convert gcode to be printable on a robot arm? I would really appreciate it! Thank you!

Not what you asked but I just had to share it I just bought the small (tiny) kit for a servo driven robot arm, the meArm. For 33£ / 50€ I simply could not resist! The plans are on thingieverse and there are also bigger 3D printed versions of it (see comments). The design of the arm seems to be rather smart, the servos / steppers are on the base not on the arm etc.

Not clue if a reprap firmware for these types of kinematic arms exist, but a GCode converter most likely. Maybe the links to the software for the meArm helps you find one. Not even sure how you call a robot arm like this? Is that still a parallel manipulator?

Great work for a minimum vitamins/parts robot. I'm humbled by the knowledge of some guys on these forums. I wish I had half of your knowledge to finish my own reprap project. Well... keep it up.

Quote
Dejay
Not what you asked but I just had to share it I just bought the small (tiny) kit for a servo driven robot arm, the meArm. For 33£ / 50€ I simply could not resist! The plans are on thingieverse and there are also bigger 3D printed versions of it (see comments). The design of the arm seems to be rather smart, the servos / steppers are on the base not on the arm etc.

Not clue if a reprap firmware for these types of kinematic arms exist, but a GCode converter most likely. Maybe the links to the software for the meArm helps you find one. Not even sure how you call a robot arm like this? Is that still a parallel manipulator?

Thanks for the reply, I've seen that robot before. I'm building another one I found on thingiverse. I'm just looking for the firmware, or a way to convert the gcode. I'll keep looking!

Well if you can program a bit you could try adapt something like marlin. Or just use the gcode interpreter in marlin.

The inverse kinematic code for the meArm is here: [github.com]

int b = sqrt ((x * x) + (z * z));        // b = distance from the origin to the start of the gripper
 //Serial.write(b);
 float q1 = atan2( x, z );                // q1 = angle between the horizontal and the line b
 //Serial.print(q1,4);
 float q2 = acos((bic_sq - for_sq + (b * b))/(2 * bicep * b)); // q1 = angle between line b and the bicep
 //Serial.print(q2,4);
 float abi = q1 + q2;                     // abi = angle between horizontal and the bicep
 //Serial.print(abi,4);
 float afo = acos((bic_sq + for_sq - (b * b))/(2 * bicep * forearm)); // afo = angle between bicep and forearm

Afaik controlling servos is simpler than steppers, so instead of generating step signals until you reach the destination over time, you just set the position you want your servo to be at:

myservo0.write(abi);           // sets the servo position according to the scaled value 
myservo3.write(afo);           // sets the servo position according to the scaled value

If you've never programmed before it might seem like a daunting task but it's not black magic. You can start simple and add more functionality as you learn. Reading code written by others is always a bit of work through. But rewarding Writing code is much more fun of course! Hope this helps.

Afaik the problem with robot arms is jerk and backlash though. Limiting acceleration of the servos (starting and ending a move slow) could help though. And servos aren't precise enough (not enough steps / mm) and not repeatable enough.

I have this crazy idea that you could improve the precision though by using computer vision to track the precise location of the effector and using some kind of algorithm to compensate for backlash and jerk and arm droop. Maybe you could use an IMU for this as well. After all our human arms are highly imprecisely manufactured as well but we "learn" to use them precisely. If you could teach a robot that then you could safe a lot on "precision manufacturing". This would also be an important step for reprap "reproducability" by creating less precise but 3D printable linkages. That's just crazy talk though

Edited 1 time(s). Last edit at 06/01/2015 04:35PM by Dejay.

Quote
Dejay
Well if you can program a bit you could try adapt something like marlin. Or just use the gcode interpreter in marlin.

The inverse kinematic code for the meArm is here: [github.com]

int b = sqrt ((x * x) + (z * z));        // b = distance from the origin to the start of the gripper
 //Serial.write(b);
 float q1 = atan2( x, z );                // q1 = angle between the horizontal and the line b
 //Serial.print(q1,4);
 float q2 = acos((bic_sq - for_sq + (b * b))/(2 * bicep * b)); // q1 = angle between line b and the bicep
 //Serial.print(q2,4);
 float abi = q1 + q2;                     // abi = angle between horizontal and the bicep
 //Serial.print(abi,4);
 float afo = acos((bic_sq + for_sq - (b * b))/(2 * bicep * forearm)); // afo = angle between bicep and forearm

Afaik controlling servos is simpler than steppers, so instead of generating step signals until you reach the destination over time, you just set the position you want your servo to be at:

myservo0.write(abi);           // sets the servo position according to the scaled value 
myservo3.write(afo);           // sets the servo position according to the scaled value

If you've never programmed before it might seem like a daunting task but it's not black magic. You can start simple and add more functionality as you learn. Reading code written by others is always a bit of work through. But rewarding Writing code is much more fun of course! Hope this helps.

Afaik the problem with robot arms is jerk and backlash though. Limiting acceleration of the servos (starting and ending a move slow) could help though. And servos aren't precise enough (not enough steps / mm) and not repeatable enough.

I have this crazy idea that you could improve the precision though by using computer vision to track the precise location of the effector and using some kind of algorithm to compensate for backlash and jerk and arm droop. Maybe you could use an IMU for this as well. After all our human arms are highly imprecisely manufactured as well but we "learn" to use them precisely. If you could teach a robot that then you could safe a lot on "precision manufacturing". This would also be an important step for reprap "reproducability" by creating less precise but 3D printable linkages. That's just crazy talk though

Thanks for the reply, I've only adjusted settings in Marlin, but have never programmed anything on my own. I will keep looking and researching. Would it be possible to input the inverse kinetic code of a similar robot in place of something that is already in an existing Marlin?
And I like your idea! I was thinking about calibrating an arm similar to how a CMM is calibrated.

I think generally you probably want to search for "arduino sketch" or program to control your robot. Not firmware. Reprap firmware are just arduino programs. Maybe this helps you to find what you are looking for. What is the name / model of the robot you want to use? And what controller do you want to use?