The sooner you get it going, and the faster it can move (as long as it can handle 64 disks), the sooner we are to the world ending. From http://en.wikipedia.org/wiki/Tower_of_Hanoi :
The puzzle was invented by the French mathematician
I had to work yesterday, so today I've been making the disks and pegs. I'm only planning on moving 4 disks because 64 would require much better batteries!
So last weekend I made the posts and disks, and glued a large washer to each disk's top. I also hot melt glued some magnets to the tip of the gripper fingers. The robot will grab by touching its magnets to the washer and lifting. It will drop by opening the gripper and breaking contact with the washer. I also ported a recursive version of tower of Hanoi to Spin and it works fine. So far so good.
But when I tried to get the robot to play a game this weekend, I hit a snag. My my kinematics look good, but there are still some lurking bugs. Basically the kinematics are internally consistent, but not aligned with the real world. So while I know the disk height and post spacing, the arms isn't lining up. So I have more debugging to do.
Martin, very nice project. If there's some depth to the hardened object to be picked up, a short sweep with ultrasonics will isolate its center position. I'm also thinking that something could be accomplished with some HALL effect sensors.
This weekend I should have some time to work on the arm. Since the repeatability is good, but in error, I'm going to treat this as a calibration problem. I'll remeasure all the constants (bones lengths, base height, servo centers, etc) that go into the inverse kinematic transform and look for any sources of error. I'll also look into the accuracy of joint angular positioning. Right now I'm converting angles to pulse width using a simple equation of lower limit, center, and upper limit. This equation assumes both accurate measurements for these three const, so I'll verify with a protractor.
If that doesn't cure things, I may need to look into possible non-linearity in the servos. If that's the case I may need multiple angle to pulse width values for each joint and linear interpolation between them. I'm really hoping servo dead band isn't to big an issue.
My laptop was also acting a bit twitchy so I backed up my files. Now it has no reason to crash and should act fine.
I'm posting my source code so other people can look at it if interested. This version of the source is much better calibrated and repeatable and I am almost ready to try Hanoi again. Finding blocks of uninterupted time is always the challenge.
Also, I didn't mention how I made the U shaped brackets. They were formed by using a rubber mallet and a block of wood. I hit the brass until it took on the desired shaped of the wood.
Hi Martin,
As I told you before (on Microcontrolled's thread) I like your robotic arm!
Your design looks simple and very strong! Congratulations! It reminds me the http://www.lynxmotion.com robotic arm design in some points!
I would like to borrow some points from your design for my next robotic arm if you don't mind.
Great work! I can also give you a design of mine for the gripper if you want (As I can see the gripper that you use it is not homemade. Isn't it?)
NikosG, feel free to borrow any ideas or the source code above. You're right that parts of that design are inspired by ideas from Lynxmotion. I'd love to see your gripper design, as I bought mine at a close out and I'm not that happy with it.
Hi Martin,
Thank you! Your code seems very organized and well-written! I think that It is an excelent paradigm about the topic of inverse kinematics. I like the way you have written your code! I send you some pictures of my gripper (screen shoots from SkethUP model). You can also get an idea from this video: http://www.youtube.com/user/NGYT40#p/u/0/TVux3GYlvMs If you find something interesting and you want more information feel free to ask me.
Personaly I have some questions about your design (e.g. I would like to know how it is adapted to the metal the back side of the servo in the second joinder).
I found a neat robot arm claw toy at TRU this weekend. The whole arm is a leadscrew. It was out of the package and not working, so the manager sold it to me for $4. I fixed it in a few minutes and it's much better than the toy claw I used in my Tower of Hanoi robot at http://www.youtube.com/watch?v=-Z8lTSX4PHs No markings on it, I can't tell who made it. It's red, blue & gray, I was guessing Transformers, but maybe not. I'll upload pics or a video later.
NikosG, thanks for the compliments on my code. Here's a picture of the other side of the servo joint which should clarify the construction:
I silver soldered a screw with a smooth section at the top to a square of brass. I used a Dremel to cut of the threaded part of the screw. I then hot melt glued this assembly to the under side of the servo. This results in a strong and stiff joint. But brass is soft and won't bind against the steel, even without lubrication.
Your gripper is similar in concept to the plastic one I am using, but looks like it can grasp larger objects.
Erco, you're taunting us with this mystery gripper. I will retaliate with pictures from my secret penguin project.
Nice photo Martin,
It explains perfect my question! You have done very detailed job! I' ll try to do this using aluminum! I love aluminum. Is my favorite material for robotics. (I am like erco with the wood )
Another point that could be improved is the way that holds the entire arm on the base. I have in mind a system with a ball bearing able to rotate the arm and simultaneously give a stability to the whole system. One of these days I'll make a new sketcup design and I'll send you the image to have your opinion.
Good efforts, guys. Martin, I have made wraparound servo brackets as you show, they definitely help reduce flex and spread the forces around. Robot arms develop high joint loads even without a payload.
For highest loads and longest servo life, I think it's best to avoid using the servo as the joint/pivot itself. Making a linkage or using the servo to drive a reduction gear allows best use of the full range of servo travel. You can get nearly 180 degrees of travel out of most servos, but any given joint travels perhaps 90 degrees or less. In those cases, a servo is better off driving through its full range of motion using a gear or linkage reduction.
NikosG, a turntable for the azimuth joint would be ideal. My joint is a work around and I used a high torque servo with metal gears and ball bearings to compensate.
erco, my first Lego arm used reduction gears with a separate pivot. It's a good idea, but adds bulk to the joint. Lego servos are continuous rotation with encoders, so I used mechanical stops and pulse counting to determine position. The servos and gears produce torque that can exceed the strength of the Technic parts, so I needed stall sensing. The whole system sank into a swamp of escalating complexity.
So for my second Lego arm I switched to RC servos, but I didn't use wrap around joints. A mistake as that arm fell over, burned down, and sank into a swamp.
Comments
Nice work! A Dremel tool could take care of the occasional solder blob and, if you like... polish the dickens out of that brass!
Amanda
I had to work yesterday, so today I've been making the disks and pegs. I'm only planning on moving 4 disks because 64 would require much better batteries!
But when I tried to get the robot to play a game this weekend, I hit a snag. My my kinematics look good, but there are still some lurking bugs. Basically the kinematics are internally consistent, but not aligned with the real world. So while I know the disk height and post spacing, the arms isn't lining up. So I have more debugging to do.
That's robotics, my friend!
Lightweight Nerf swords, but cool nonetheless!
If that doesn't cure things, I may need to look into possible non-linearity in the servos. If that's the case I may need multiple angle to pulse width values for each joint and linear interpolation between them. I'm really hoping servo dead band isn't to big an issue.
My laptop was also acting a bit twitchy so I backed up my files. Now it has no reason to crash and should act fine.
Also, I didn't mention how I made the U shaped brackets. They were formed by using a rubber mallet and a block of wood. I hit the brass until it took on the desired shaped of the wood.
As I told you before (on Microcontrolled's thread) I like your robotic arm!
Your design looks simple and very strong! Congratulations! It reminds me the http://www.lynxmotion.com robotic arm design in some points!
I would like to borrow some points from your design for my next robotic arm if you don't mind.
Great work! I can also give you a design of mine for the gripper if you want (As I can see the gripper that you use it is not homemade. Isn't it?)
Thank you! Your code seems very organized and well-written! I think that It is an excelent paradigm about the topic of inverse kinematics. I like the way you have written your code! I send you some pictures of my gripper (screen shoots from SkethUP model). You can also get an idea from this video: http://www.youtube.com/user/NGYT40#p/u/0/TVux3GYlvMs If you find something interesting and you want more information feel free to ask me.
Personaly I have some questions about your design (e.g. I would like to know how it is adapted to the metal the back side of the servo in the second joinder).
I silver soldered a screw with a smooth section at the top to a square of brass. I used a Dremel to cut of the threaded part of the screw. I then hot melt glued this assembly to the under side of the servo. This results in a strong and stiff joint. But brass is soft and won't bind against the steel, even without lubrication.
Your gripper is similar in concept to the plastic one I am using, but looks like it can grasp larger objects.
Erco, you're taunting us with this mystery gripper. I will retaliate with pictures from my secret penguin project.
It explains perfect my question! You have done very detailed job! I' ll try to do this using aluminum! I love aluminum. Is my favorite material for robotics. (I am like erco with the wood
Another point that could be improved is the way that holds the entire arm on the base. I have in mind a system with a ball bearing able to rotate the arm and simultaneously give a stability to the whole system. One of these days I'll make a new sketcup design and I'll send you the image to have your opinion.
For highest loads and longest servo life, I think it's best to avoid using the servo as the joint/pivot itself. Making a linkage or using the servo to drive a reduction gear allows best use of the full range of servo travel. You can get nearly 180 degrees of travel out of most servos, but any given joint travels perhaps 90 degrees or less. In those cases, a servo is better off driving through its full range of motion using a gear or linkage reduction.
erco, my first Lego arm used reduction gears with a separate pivot. It's a good idea, but adds bulk to the joint. Lego servos are continuous rotation with encoders, so I used mechanical stops and pulse counting to determine position. The servos and gears produce torque that can exceed the strength of the Technic parts, so I needed stall sensing. The whole system sank into a swamp of escalating complexity.
So for my second Lego arm I switched to RC servos, but I didn't use wrap around joints. A mistake as that arm fell over, burned down, and sank into a swamp.
But the third arm! The third arm stayed up!