This instructable is a mechanical arm actuated by antagonistic elastic muscle pairs. It is intended for exploring how living things learn to move. Unlike conventional rigid robots, living things learn to control a flexible body with elastic muscles capable of fast energy-conserving "ballistic" gestures. This is a passive prototype of the mechanical elements. It is not yet extended to include motors, sensors or a "useful" end for grasping or drawing (drawing is my particular interest). In this demonstration design, just to get a feel for how the arm moves, the muscles are literally lengths of elastic without any electrical components. It is designed to be played with manually in order to get a feel for how an arm controlled by elastic muscle pairs "wants" to move.

Step 1: Materials

Here's more or less what I used to make the first iteration of this arm. You might have better ideas. If you do; please let me know.
  • 720 6mm stainless steel ball bearing (I recommend getting plenty more in-case some roll away)
  • Two 5x420x720mm and one 2x420x720mm sheets of clear acrylic
  • Five 500mm lengths of 6mm stainless steel screw thread (and about 30 6mm bolts).
  • About 100 18mm long 5mm thick bolts and nuts (I recommend checking whether the bolts you use leave enough clearance for the bearings to roll free. I ended up using countersunk bolts and drilling a countersink into the bolt holes where need by).
  • Strong fishing line
  • A couple of meters of elastic resistive exercise tube

Step 2: Design

The arm has three joints, all in the same axis, each pulled by a pair of elasticated control-lines. Having all the joints in the same axis might sound limited, with one of them even seeming redundant for the conventional job of reaching. This arm is intended for learning how to coordinate fast gestures using pairs of elastic muscles and not necessarily doing useful things like picking something up.

Central to the design is a method to laser cut bearings. Each bearing contains 3 rings of stainless steel ball-bearings. There is a central ring of 10 ball bearings and an upper and lower ring of 15 ball bearings each. There are a total of 18 of these bearing units in this arm functioning as both pivots and control-line guides. They're not nearly as efficient as ready-made sealed bearings. The advantages are that you can design and make them yourself and modify them to be a hybrid of control-line guide and structural element.

Behind the shoulder joint is a long extent for the elastic part of the control-lines. This is the area that could be modified in future to include any manner of motors able to automatically pull the 6 control-lines.

Step 3: Cutting

Here are PDF files for the three 720x420mm sheets that the robot is cut from. Two are 5mm deep and one 2mm deep. You'll probably have to change the line color and thickness to suit whichever laser cutter you have access to. For example the laser cutter I used only cuts lines that are fully saturated red (in RGB, not CMYK format) and less than 0.01mm thick.

Handy Hint: You might find you're missing or have broken a component and need to cut a replacement. If you change the color of all the lines that have been already cut, then you can place new shapes within the sheets you've already cut. I save a new copy of the cutting PDFs each time I cut an additional component.

Step 4: Assembly

Start by securing a scaffolding of 5 vertical 6mm screw threads to a sturdy base. I bolted the threads to a wooden base. Next is the laborious process of following the plans to gradually assemble all the laser cut components and ball-bearings. With the current design I found inserting the ball bearings to be very fiddly. I ended up having to squeeze the lower rings of ball-bearings into place one by one from below with needle nosed pliers.

Handy Hint: When mistakes happen I found a big magnet really useful for rescuing misplaced ball-bearings.

Bolt the components together as your build. I found it useful to temporarily hold the vertical stacks of bearings in place in nuts on the scaffolding of vertical screw threads. This involves a lot of spinning nuts up and down the screw threads as each bearing is constructed.

Handy Hint: I found a handheld power-tool with a polishing attachment really useful for quickly spinning nuts up and down the vertical screw threads. I used a Dremel (eg: amazon.co.uk/Dremel-Rotary-Tool-Series-Accessories/dp/B000KJRQN0 ) on its lowest power setting.

When all the components are bolted together in place cut away the screw thread scaffolding supporting the two arm joints (I used the Dremel again, Caution: the threaded bolts get very hot when cutting, don't get burnt, and if you smell acrylic starting to melt then stop and let it cool down).

Step 5: Rigging Control-lines and Muscles

This diagram describes how I routed fluorocarbon fishing line through the arm. Each control-line is secured to one side of the joint bearing at which it ends. I secured it by looping it around a bolt that holds the bearing before tightening the bolts (I had a problem with the lines eventually managing to wedge between the bolted components so I suggest bolting very tightly or finding a better solution).

Each control-line loops once around each bearing it passes. The loops prevent each line from leaving its guide no matter how the arm bends. The lines alternate between looping clockwise and anticlockwise. This criss-crossing seems to help isolate the effect of each muscle to one joint. Without the criss-cross arrangement the whole arm tends to bend one way or the other and is less able to pull joints in opposite directions for an s-shaped reach.

The rear of each control line is tied to a length of elastic. Knot the lines at a length where the elastic cannot go slack when the arm is bent, and so that the muscles are all the same length when holding the arm pointing straight ahead.

There you have it. A lovely looking bendy arm. The next iteration will refine the design and extend it to include motors of some sort. Let me know if you build the arm and let's put our heads together to build it better.
Impressive! I could see how this could "evolve" into something even more complex and life-like.
Have you considered applying this to an exoskeleton like set up to allow control of joints where a servo would not be an efficient set up? I am thinking something like this might work in my project

About This Instructable




Bio: I grew up playing with computer programming, with my first software title being published at the age of 17. Following a degree in Architecture at ... More »
More by edburtondotnet:Elastic-muscled mechanical arm with 3 muscle-pairs, 3 laser-cut joints and 18 bearings 
Add instructable to: