Here’s a 1 year anniversary post on my electric muscle research.
I started a year ago as a for-fun project in my spare time. And now it’s grown to a full-time focus.
My goal is to make the best biomimetic robots in the world. Won’t rest until I get there!
Good progress on the robotic muscle! In this clip it’s lifting 50 grams.
This actuator design has never really been explored before. The theoretical cap is 10lbs/cm^2, which means I’m at 1% of the possible strength.
Goal right now is to make it stronger!
Super great book to read. Especially for people interested in microfab.
Has everything from glassblowing, to laser optics, to vacuums, etc..
But most important thing is that I can open at any page, and find it fun to read.
Quick update on the actuator!
Main update is I'm testing a material that gets me 10x strength compared to before.
Imagine the last clip I posted, but 10x stronger. Would be wild! More details about what's new in thread.
Making progress on the electrical muscles!
There’s a lot of room for improvement, but this is the bare bones V1 of what’s possible.
Next I gotta
- decrease the layer thickness
- change up the material
- fine-tune the folds
- optimize the control waveform (bipolar).
Last year
- went from knowing nothing in electrostatics, to being at the forefront of the field.
- experimented on half a dozen variations of the electric muscle actuator.
- learned a crazy amount about high voltage electronics.
- incorporated my first company w/ stripe atlas.
Ok pretty cool progress! Now it’s lifting weights at a pretty low voltage.
Another 10x improvement is easily possible. Like in the next few days.
Don’t be surprised when it starts lifting 1kg weights soon!
Update on the robotic muscle project.
The electrostatic actuator is running at super low power consumption (~0.1 watts)
It has high frequency control, and high precision.
Really excited by the progress! A lot more coming next month.
We simply don’t have an actuator technology that’s comparable to human muscle. There is no artificial muscle that’s even close.
BLDC - too heavy, not enough torque without a gear reduction.
Hydraulic - hard to back drive without using leaky loose seals, and requires a pump, lots
A pretty dramatic increase in force! New material makes a huge difference.
There’s nothing holding the copper electrodes together besides electrostatic force.
I feel like Edison, sourcing all these different materials, learning by trial and error.
You know to be honest.. I’m not entirely sure what’s happening here.
Why is the alligator clip visibly lifted up, but the actuator doesn’t compress??
Really cool outcome nevertheless.
In the coming year I want to
- further improve the muscle performance.
- build a robot v1 that demonstrates superior biomechanics.
- have my first investors in the company. looking to raise 50k from angels.
In terms of go-to-market, I think there’s a really strong case for making incredible animatronic robots for the entertainment industry (Disney et al.).
Robots today move super stiff, and if the electric muscles can lead to significantly better robots, that’s a huge deal.
Did a ton of soldering work today. Learned a lot and it was fun!
It’s a HV power supply. I’ll be using it for my project on artificial muscles.
Have not added the HV components yet, but was able to communicate with the microcontroller no problem.
Ran some more experiments today! The actuator is getting better.
Still running it at a really low voltage. The force output is proportional to the square of the voltage, so running it at high-voltage should be extraordinary.
Let’s see if I can make that happen tomorrow!
There’s still a huge way to go in terms of optimizing the muscle design and the power electronics. Lots of room for improvements.
Improving by a first 10x is relatively easy, and a further 10x should be possible. That’s 2 orders of magnitude stronger than where it’s at now.
Same prototype as the video yesterday, but now it’s hanging!
I’m still facing an issue with arcing (since the copper is exposed). Next design is fully insulated, so should be able to run 3-4x higher voltage.
In theory should result 9-16x the strength. But that’s just a theory..
New tests for a new year (!)
Wish me luck as I upgrade my power supply, to avoid charge retention.
Still, good progress.
Also, there’s a fancy machine at Waterloo that I want to try out. Can prototype flex pcbs in a few hours, don’t have to wait 7 days for china.
July 2022 Update:
- I'm starting a new project tackling the problem
@paulg
describes in "The Lesson To Unlearn".
- I worked at
@Primer
figuring out how to teach math as play.
- I made a math video in the style of
@3blue1brown
explaining group theory.
I'm in a period where I'm reinventing myself, and it felt like a hard (but correct) decision.
I feel like this becomes a much harder decision after one achieves more tangible success.
Made a fun physics demo! The electrostatic force is really strong.
The two strips totally stick together when there’s a voltage applied.
This is due to charges accumulating at the surface.
Vapor deposition is incredibly underrated. If I had 100$ million in grants, I would spend it all doing experiments in vapor deposition.
Chemical vapor deposition. Atomic layer deposition. Laser chemical vapor deposition. Electron beam induced deposition. Focused ion beam depo--
Very slight compression visible. I’m having a problem that the hinges are far too rigid. Working on fixing it!
For a 2g actuator lifting up a 20g weight, that’s still pretty good.
Very interesting to me that both
@patrickc
and
@LauraDeming
posts on advice start with the same thing:
"Go deep on things. Become an expert."
"Get obsessed with a topic"
Learning to play the piano but I am NOT using sheet music. Doing it by ear.
So far I’ve learned:
- melody for Rachmaninov concerto No. 3 at 4-6 mins
- Satie’s Gymnopédie No. 1
- Celeste Resurrections
- Debussy’s La fille aux cheveux de lin
Running some backyard oven experiments.
The test is: can I weld two plastic films together, and what is the weld strength?
This is one option for how I’m going to manufacture the folding strips for the actuator.
Did some laser micro-machining to prototype a new Flex PCB design for the robotic muscle project.
Being able to iterate daily rather than every week will be huge.
Nov 2022 Update
- I'm back in Toronto and I'm doing research into 3d printing
- Looking at vibratory fluidization, which is a method for making powders behave like fluids
- Designing an initial prototype to test out my hypothesis
If you're a person that does stuff without needing external approval, please consider applying to Interact. I have such a high trust in the community. Everybody I meet are people that I can just *relate* to and it's incredible.
I’m testing the electrostatic actuator I built at 1000V.
The more I shrink the design, the better it scales. Halving the distance between electrodes is 4x the force output.
The next iteration is going to be much better.
I've been doing research into fluidization and how it could be used to make a 3d printer that's a lot faster than the ones that exist today.
Ran a quick experiment to validate some research, and recorded a video to explain. Check it out!
My curse, but also my blessing, is that I'm only interested in projects with asymmetric upside.
Every time the project upside becomes capped, I lose interest.
Always hunting those positive black swans.
Right before I fall asleep, I come up with a good idea I should do. I wake up, do that thing, and get ready to fall asleep. But then I get another good idea I should do, and the cycle continues. And now it’s 3:30AM.
I've been writing one explanation a day for 2022. Here's the details:
- On average I write 700 words per day.
- On Jan 7th I wrote 3000 words.
- I say “tensor” 43 times.
- I wrote nearly 20 000 words so far.
Please check out the meta post here:
An update from me!
- Sustainably continuing to work on "learn by explaining"
- Learning basic math and making it addicting for myself
- Making YouTube math video on group theory
- Hosted learning retreat in Muskoka
Listen to me explain it here
@pronounced_kyle
True. But power law means that the 1% dominates real world impact. It’s mostly non-academic.
- steam engine
- wireless telegraphy
- vulcanized rubber
- dynamite
- powered flight
So, that's the high level update. For those that are interested, here's a more technical explanation of what's going on.
What's happened since my post last month is that I've tested lots of film materials for their electrostatic strength.
The experimental set up is quite
December 2021 update:
I'm pretty sure I'm on a good path, but I'm always anxious because of the inherent uncertainty of not knowing where I will end up.
If you want to hear me explain it, feel free to listen to me talk in this video!
I always come back to this experience I had playing volleyball where for a full year, I'd been playing with intermediates players, and got slowly better.
Then one day I joined the advanced group, and improved in 1 week the same amount as the whole last year.
Stumbled on an old application from 2 years ago, and I think I was more idiosyncratic back then compared to now.
I was super fixated on curing aging computationally, which was maybe naive, maybe not, I don't know.
But I think the impassioned self-belief I had was a good thing.
In 1 month I'm probably going to understand symplectic manifolds. Am currently struggling to understand it but I know I'll make a breakthrough soon enough. It's the same feeling I had with topology, smooth manifolds, tensors, tangent spaces, tensor fields
I should never let myself be spoon fed information.
My instinct when I get stuck is to turn to textbooks or courses to build a "knowledge base".
This passive learning is bad -- my mindset should always be asking questions and figuring things out.
March 2023 Update
- Research into 3d printing power semiconductors for electric vehicles, drones, and the electric grid
- Recorded a video talking about all of my failures from the last 2 years (and some success!)
- Visited Austin and Boston
Listen to my monthly update below!
TLDR:
- Started the Quantum Mafia
- Building out SkillTree
- Making projects in quantum simulations of chemistry
- Moving back to Toronto
YouTube Link:
It’s quite possible that top-down learning is currently better than bottom-up only because of arbitrage reasons. Nobody is doing top-down, everybody is doing bottom-up. Hence it’s much easier to come up with novel insights when doing top-down.
What I'm exploring: additive manufacturing by vapor deposition.
This would open up a whole new market by enabling 3D metal printing to reach serial production.
I explain the main technical gaps that need to be addressed for this process to be viable:
I feel like injection mold tooling is very underappreciated for just how big of an impact it has on manufacturing and our daily lives.
Corollary to this is that I think what
@aphysicist
and
@austinbishop
are doing at Atomic Industries is super underrated.
Self learning math.
You get recommended courses and textbooks.
You don't end up listening to them.
That means sometimes missing out on amazing content.
Here's my solution to the problem:
I’m fine tuning the rigidity of the actuator, and haven’t solved that fully yet. But making good progress!
I’m getting all sorts of wacky results like the one in this clip.
Excited to see where it goes!
I'm assembling a working actuator with this new material ASAP.
By the way, order of magnitude improvements are not surprising here, since this tech is so underexplored.
Just look at my previous actuators to see the progress. I've learned a lot since 6 months ago.
What if we use AI to take scientific papers and translate them down to the pure empirical results.
Cut out the theory, the simulations, the models. All the fragile stuff. Leave me with just the results.
Looking into the future, there's this short term 10x strength increase I'm working on, and then I think I can probably 10x it again after.
This would bring the actuator strength to 10lbs/cm^2, which is equivalent to human muscle.
Understanding theoretical physics from first principles is so much fun. It goes like this:
logic -> set theory -> topology -> topological manifolds -> differentiable manifolds -> bundles -> symplectic/metric geometry -> modern physics!
@MajmudarAdam
For the fastest turnaround time, it could be interesting to use the semi equipment from a local university!
It’s a fun way to get trained on the machines.
Access cost is typically around 50$/h for individuals.
“I certainly have a routine, but the most important thing, when I look back over my career, has been the ability to change routines”
- Anne Rice
From reading daily rituals!
@AviSchiffmann
totally agree. it's a barbell strategy.
1) casual reading, with no effort at all of remembering things, and letting is subconsciously load intuition
2) high energy, figure shit out, read highly technical papers and documents. challenge the concepts. etc.
nothing in the middle!
Currently: I'm exploring this technique called Electron beam induced deposition, that could potentially be used to 3d print multi-material structures on a nanometer scale.
Great interview by
@AlexMasmej
in French.
You know, French is my first language, but it's crazy how dominant English is in tech.
It's a power law, but for languages. All but guarantees that English will cement itself as the new Lingua Franca.