Most humanoid robots are constrained by having to use rigid actuators like electric servos. They are immediately prevented from investigating and using most of the biomechanical tricks that human (and animals) use in order to move around.
I am building a robot that embraces a number of such tricks and I will utilise custom soft robotic actuators to give this robot a large range of movement.
The robot will use a range of materials ranging from rigid (bones), flexible but inelastic (fascia, tendons), flexible and elastic (ligaments) to Flexible and actuated (muscles). Joints will not be rigidly constrained, but instead be constrained using the range of flexible tissue types.
I will also post updates on my anthromod website and youtube channel.
So far I have run several tests, including printing part of a skeleton and casting custom passive silicone ligaments, in order to test the types of joints that can be made. Actuator tests will follow soon.
Since the hand welded method was time consuming and inaccurate I needed something better. I ordered a cheap Neje Master '7W' laser engraver from China and started some experiments.
Originally I planned to cut the paper and then seal the actuators with the standard way with an iron. However a quick test showed me that welding the plastic together was more than possible, and so I've been refining that process since.
So far I've found that it's best for the plastic to be placed on top of black card, and then sandwiched between 2 planes of glass. The glass I used were borosilicate 3d printer beds. A weight can be added to help hold the glass flat. The sous-vide pouch I was using for the glass wasn't sealed perfectly flat around the edges and had some small undulations to them. This lead to extra space between the glass sheets. I decided to cut all but a corner of the pouch weld off. This held the plastic sides together but without the undulation.
Of course the Neje software didn't like the modern dxf formats that Fusion 360 output. So I used Librecad to convert the modern dxf files to 2007 versions.
So far I've found that the laser settings (450nm 2500mW optical power) that are best suited for welding are 60% power 1ms dwell time. The lines in the dxf file are best placed 0.075mm apart, although this is based on the rastering settings rather than extensive tests.
I made a new 3d printed connector. I decided that it was best to use epoxy to seal the connector in. Whilst I doubt the bond between the epoxy and the polyethylene plastic is strong, I designed it so the epoxy forms a plug and the geometry of the weld line prevents the plug from popping out. So far this has worked up to 12psi and I've had no failures so far.
The next step will involve various muscle geometries, including sandwiching layers together so that they can provide greater than 25% contraction ratios.
So as things have started to straighten out I've been able to get on with some experimental designs over the last couple of weeks.
I've started work on a film based muscle that's a bit like the Peano muscle except pneumatically driven.
After a few attempts with various polyethylene based sheets I ended up trying Sous-vide cooking pouches. These work quite well as they are designed for bonding together, no tiny pinholes, and are intended to resist pressure. They are also a composite of a nylon exterior and a polyethylene interior. This means I can bond the interior without (too much) risk of puncturing the exterior and damaging the muscle.
Whilst I can seal the pouches with a clothes iron, or a soldering iron I needed a way of keeping some areas of the muscle from bonding. These areas would form the air chambers. I tried normal paper and it worked well in preventing some areas from bonding, the paper was then difficult to remove. Whilst the paper might not have affected the movement of the muscle much, it was an obstacle to airflow through the muscle.
I looked around for a solution when it dawned on me that edible paper was a thing. I found that paper based on potato starch was particularly easy to dissolve (well weaken and tear easily) in water. I've tried the potato paper method and it worked ok. The sheets I had were a little thick and caused a few tiny tears where I had stretched the top plastic layer too much when pushing with the soldering iron. But these were simple to fix.
I used a drag knife on a 3d printer to try and cut the paper, but the drag knife method often caught the paper or plastic I was trying to cut. I've ordered a small laser engraver that should be sufficient to cut the paper, and perhaps weld and cut the plastic directly. This should allow me to build smaller muscles, which would be a lot of use in actuating hands or faces.
I did a few tests that just used 'hand drawn' welds using a soldering iron. These started as pressure tests, for the 3d printed pneumatic connectors I was using. But I moved on to subdividing them into smaller channels and investigating how they buckled under pressure.
I then moved on to combining 3 muscles together in a manner inspired by the deltoid muscle group. This worked to show that multiple muscles could be constructed on a single pouch, and then stay connected, whilst being folded into a 3 dimensional structure. Think of it as a robotic Pepakura model.
My last experiment was actually a software simulation of the film muscles in Blender 2.82. This update improved the cloth simulation inside Blender so that it could model pressure within an enclosed cloth volume. I don't expect great accuracy, or the ability to model forces well, but it might be useful to get an idea of amount of contraction possible and the sort of buckling that might happen.