Why Not Teach a Robot to Play a Kalimba

Especially a robot with the exceptional precision & control of the Mecademic 500.

This was just a short trial run, me basically playing around to see if:

  1. It was possible to create a robotic motion which played or “plucked” the tines to produce a tone similar to what is made when a human finger plays the keys, and
  2. The movement of the robot with “safe” position between tines / keys would allow for reasonable pace / rhythm in a short practice tune.

The first kalimba I tried is a nice basic unit with lovely tone, it’s been a lot of fun as an introduction to the instrument.

You can check it out on Amazon here: 17 Key Kalimba

I found that, though not a problem for humans, the tines required a bit too much force for the robot end effector. The 1.5mm aluminum probe tip kept bending after a few keystrokes.

A little searching online found another type of kalimba, this one with a specialized key design that required a lighter touch. It also came in more than 1 variety, with different numbers of tines.

This kalimba is more polished / professional; I purchased one from Amazon with 2 day delivery: MOOZICA 36 Keys Chromatic Kalimba

There wasn’t a ready-made .stp file or any CAD available for the kalimba, so I improvised in AutoCad Fusion to make a basic stage layout:

The tines are smaller and more rounded than the first kalimba, which made precision end effector placement & motion critical to tone. Just moving the EOAT vertically in a short rapid motion didn’t work; there needed to be a diagonal component, effectively “sliding” the probe tip off of the tine in a particular manner.

The robot motion was a sequence of move-to positions with some speed changes and delays to produce the timing of the tune. The move-to-safe motion is larger than it really needs to be, but I was concerned about clearances as I moved the blending values higher for smoother / more rapid motion. This is the code I programmed for the short video sequence:

I am working on a better CAD model of a kalimba, with the tines placed accurately. The whole stage could then be imported into a simulator like RoboDK to teach songs or melodies quickly. I did this one by eye, experimenting with different positions & motions to produce the best tones.

You can watch the YouTube video here: Mecademic Meca 500 Kalimba Concert 1

Product Links:

17 Key Kalimba
36 Key Kalimba

Robot stage

Optical Plate Flat Aluminum Honeycomb Breadboard: https://amzn.to/4aKMuYR

3D Printer

We highly recommend the Flashforge line of 3D printers for printing your robotic EOAT and staging parts:
FLASHFORGE 3D Printer Guider II Large Size Intelligent Industrial Grade 3D Printer https://amzn.to/3VKaGGA

FLASHFORGE Guider IIS 3D Printer Auto Leveling with High Temperature Nozzle https://amzn.to/3xpxJMF

FLASHFORGE Adventurer 5M Pro 3D Printer with 1 Click Auto Printing System https://amzn.to/3TQxYYz

FLASHFORGE Adventurer 5M 3D Printer with Fully Auto Leveling
https://amzn.to/3TMprpS

FLASHFORGE 3D Printer Adventurer Series with Auto Leveling with Quick Removable Nozzle https://amzn.to/3PSmeDO

Disclaimer: As a participant in the Amazon Affiliate program, we earn a small commission if you make a purchase through one of these links, at no additional cost to you.

The Hurling Video

Playing around with TCP settings on the Mecademic one day, and this morphed into the thought that it would be fun to have a robotic sporting match to showcase this very useful function.

Basketball was out (couldn’t figure out how to dribble the ball repeatably), and soccer didn’t seem practical given the lack of legs in the participants. Irish Hurling came to mind while contemplating ice hockey (the hurley and the hockey stick have vaguely similar shapes).

Ok, so we have the sport identified, now what about props… The first time I watched a hurling match, I was sure the hurleys must have a scoop-type feature to keep the sliotar in place as the players raced up and down the field. The face of the hurley is in fact NOT concave, it’s the skill of the hurlers (camogie-ers?) and the slightly raised stitching on the sliotar that keeps it in place. It’s as if the players have their own TCP functions to keep the center point firmly in the middle of the flattened end of the hurley!

Here’s a great YouTube video from GAA MAN with players demonstrating their hurling skills: https://youtu.be/44Gi2IarBiI

Amazingly enough, miniature hurleys and sliotars aren’t for sale on Amazon or Etsy (and I did a lot of searching). Looking at some online images, I designed a model hurley in AutoDesk 360 that at least keeps the shape, if not proportions, of the real thing. Here’s a great intro to AutoDesk Fusion 360 if you’re not already a user: AutoDesk Fusion 360

I ended up printing the hurleys both ways; in the video, the Mecademic’s hurley is flat while Neds and the xArm’s have one flat, and one concave side.

Autodesk 360 image of printable hurley, ~ 6" overall length

2 printed hurleys on the model playing field

The hurley strapping is electrical tape.

OK, now for the sliotar… I spent an inordinate amount of time trying to stitch a tiny canvas ball, but the material shredded in places. Next, I cut up an old pair of leather gloves and tried sewing a leather one using a tiny baseball pattern. It turned out more square than spherical:

Result of attempt to sew tiny leather sliotar

Ned actually liked this one best; it didn’t roll at all, so was easier to hold on the hurley.

The final sliotar was a design based on some historical sliotar images on the National Museum of Ireland’s website. They were made of cow or horse hair. A ball made of coarse twine seemed about as close as I was going to get; once the glue set, it was hard and rolled reasonably well.

Coarse twine sliotar

The goal was cobbled together using twine and mesh netting on a wooden frame.

Programming the winning goal motion on the Mecademic was a challenge; suffice to say, acceleration settings were key. Once the proper settings were found, though, Mike made the goal in ~ 75% of the shots. Here’s the final program used in the video:

You can watch the Robofiesta hurling video here.

Getting to Grips with Grippers

In many automated assembly applications, a well-thought-out custom gripper design is essential. In this article I’ll talk about the process and iterations needed to accomplish the Meca 500 Diode Load sequence in RoboFiesta’s Mecademic Meca 500 Diode Loading Demo on YouTube.

Some of the things you’ll need to consider are:

  • What’s being gripped / picked. Dimensions, fragility, prestage setup
  • Speed required for the application
  • Precision / repeatability required for the application
  • Incoming material variability
  • Placement location with regards to EOAT orientation
EOAT will change depending on material being handled, and pick/place location considerations

For a part in a given location / orientation- a diode mounted on a breadboard, for instance- the gripper design would need to change depending on which direction the gripper fingers extend from the gripper mechanism.

It’s a good idea to do some dry runs with mock material to determine the requirements. For the diode handling, we could have picked with the gripper fingers extending in either X or Z from the mechanical gripper body mounted on the robot flange:

Which directions the gripper fingers extend also affects the toolpath you will program. In our diode sequence, the gripper fingers extend down in X from gripper body.

It took several iterations of grippers to develop a design that would firmly grasp the slightly tapered diode body without slipping. The final design featured a small relief or notch feature to accommodate a small lip on the bottom of the diode body.

Gripper with relief feature for diode lip

Finally, having a capable 3D printer like the Flashforge Creator Pro is invaluable for on-the-fly iterations for functionality and changes in setup orientation of your grippers. I’m currently using 1.75mm ABS Transparent filament, it’s a great material for general purpose applications.

Watch a timelapse of a gripper print here.

There are other factors to take into consideration depending on the situation, but those are the basics I start with when making new gripper designs.