An artificially intelligent, open-source bionic leg(qz.com)
qz.com
An artificially intelligent, open-source bionic leg
https://qz.com/1636413/an-open-source-ai-bionic-leg-is-the-future-of-prosthetics/
31 comments
This leg uses advanced AI and Machine Learning algorithms with a cloud Prosthetic as a Service (PaaS) model. Utilizing the latest in state of the art open source blockchain technology, this leg will bring added value while synergizing with your efficiency to optimize your bottom line.
The true cost of a digital currency in the energy required to mine new blocks of currency. By using just a fraction of the energy naturally generated through the movement of the leg while walking, each step could bring you more wealth in STRIDE coin(™).
STRIDE - The future is just a step away.
STRIDE - The future is just a step away.
You forget there is an option to run on-prem
Actually, I'd be more interested in one that can run in the cloud...
Not enough digital transformation
it's a win-win-win
I'm the owner of a very dodgy right leg (fibula hemimelia) that will probably only last another 10-20 years. 15-yr-old me dreamed of being able to eventually replace it with something more iRobot than steel-post.
Godspeed to these amazing researchers.
Godspeed to these amazing researchers.
More than any other field prosthetics MUST be open source. If you do not have full control of your body do you really have control of anything?
Breakdown of the cost:
* Star Prototype (machined parts) ……………….. ~$15.6 k
* Dephy (actuators and encoders) ……………….. ~$8.7 k
* SDPSI (belts) ……………………………………… ~$0.2 k
* McMaster-Carr (bearings, fasteners, etc.) …….. ~$1.6 k
* Misumi (shafts) …………………………………… ~$0.1 k
* Motion Industries (angular bearings) ………….. ~$0.8 k
* Amazon (batteries and case) …………………… ~$0.5 k
* K&J (magnets) …………………………………… ~$0.0 k
* Shipping charges ……………………………….. ~$0.9 k
* Total Cost ……………………………………….. ~$28.5 k
The greatest cost (~$15.6k) is for the machined pieces.
That seems seriously inflated though.
Wonder how much someone with a well calibrated workshop CNC could get that down to...?
* Star Prototype (machined parts) ……………….. ~$15.6 k
* Dephy (actuators and encoders) ……………….. ~$8.7 k
* SDPSI (belts) ……………………………………… ~$0.2 k
* McMaster-Carr (bearings, fasteners, etc.) …….. ~$1.6 k
* Misumi (shafts) …………………………………… ~$0.1 k
* Motion Industries (angular bearings) ………….. ~$0.8 k
* Amazon (batteries and case) …………………… ~$0.5 k
* K&J (magnets) …………………………………… ~$0.0 k
* Shipping charges ……………………………….. ~$0.9 k
* Total Cost ……………………………………….. ~$28.5 k
The greatest cost (~$15.6k) is for the machined pieces.
That seems seriously inflated though.
Wonder how much someone with a well calibrated workshop CNC could get that down to...?
We have a full CNC shop in house with the latest equipment. I took a quick look at the design. It really isn't optimized to reduce fabrication cost. The CNC work on this leg is substantial and very expensive. It simply isn't designed for low cost manufacturing. I am not being critical, just stating a fact in response to your question.
In CNC machining cost reduction can be obtained through the reduction of what's called "setups" (how many sides of a part you have to machine), judicious selection of feature types (some types of cuts require the machine to slow down a bunch, special tooling or both) and fixtures.
The last one, fixtures, are custom made tools to hold the raw material and in-progress machined parts for machining. They can make a significant difference in cost. Without them a machinist has to spend a lot of time manually managing what is known as "work-holding" (how you hold the material for machining).
Beyond that, it's a matter of volume. If you are only making one set at a time, without tooling and without optimization to reduce CNC machine cycle time and tooling requirements it is going to cost a lot. If, on the other hand, the design is optimized for the process, part counts are reduced, setups are reduced and you pay for palletization tooling (allows you to make more than one part at a time) cost can be reduced substantially.
In addition to this if you can reduce or modify the size and shape of the parts such that you can fit, say, ten of them within the working envelope of a machine rather than five, it can make a huge difference in cost. This, again, requires the design of a palletizing fixture.
Fixturing and workholding for production of a few parts at a time costs money. Think of it this way: Each part you have to make requires a fixture to be made for every setup. In other words, if you machine from the top, bottom, left and right of an aluminum block and want to design a fixture to hold ten of them at a time, you have to design one fixture for each setup or side you are machining. There are ways to reduce this by working closely with the designers and optimizing for work holding.
You can also integrate work-holding such that one pallet can hold the part in different setups. In other words, you hold four parts in setup 1, four in setup 2, four in setup 3 and four in setup 4. A pallet designed like this produces four finished parts per machining cycle. After each run you move the parts to the next setup location. So, the four parts that were being machined from the top are moved to the right one position and clamped onto an area of the fixture designed to hold them for bottom machining. The other parts are similarly moved one position for the next setup and the parts on the last setup slots come out of the machine for finishing (clean, debur, inspect, post-process).
If you do the math on this the scope of the manufacturing problem starts to be revealed. If you have ten parts to machine, and each part requires, on average, machining from two sides, you need to design twenty individual holding fixtures. Some can be integrated into a unified multi-setup fixture but the design work isn't any different, in fact, it can be more expensive. If the parts require machining from four sides you need 40 fixtures.
Once you are done with mechanical design of the fixtures and setups each one of them requires CNC programming and tool selection. This is a detailed process that requires time, expertise and testing, lots of all of them. In a design like that of this leg the design, fabrication, programming and testing of fixtures can easily cost much more than what the posted cost for machining one set of parts.
In short, cost reduction for something like this requires detailed and intense engagement with the manufacturing process, which would lead to design changes necessary in order to reduce manufacturing complexity, machine cycle time, fixture count and complexity, material waste, etc.
One way they could have reduced the cost to manufacture these for other people could have been to, for example, pay for manufacturing fixtures to make four to ten of these at a time. Customer #2 would then benefit from this and possibly cut their cost by a substantial amount (25% to 50%). At the limit the cost of machined parts become asymptotic with material cost + hours on the machine + aggregated costs-per-hour for the shop (people, wear and tear, power, etc.).
It's an interesting project. If the market is large enough it might make sense for someone to pay for the design and fabrication of the tooling to offer these parts at a lower price.
At a larger scale you'd have to design the parts for a casting + machining process, which will make them far less costly.
Metal 3D printing might be another option but that's not as cheap as one might think and has its own design requirements.
In CNC machining cost reduction can be obtained through the reduction of what's called "setups" (how many sides of a part you have to machine), judicious selection of feature types (some types of cuts require the machine to slow down a bunch, special tooling or both) and fixtures.
The last one, fixtures, are custom made tools to hold the raw material and in-progress machined parts for machining. They can make a significant difference in cost. Without them a machinist has to spend a lot of time manually managing what is known as "work-holding" (how you hold the material for machining).
Beyond that, it's a matter of volume. If you are only making one set at a time, without tooling and without optimization to reduce CNC machine cycle time and tooling requirements it is going to cost a lot. If, on the other hand, the design is optimized for the process, part counts are reduced, setups are reduced and you pay for palletization tooling (allows you to make more than one part at a time) cost can be reduced substantially.
In addition to this if you can reduce or modify the size and shape of the parts such that you can fit, say, ten of them within the working envelope of a machine rather than five, it can make a huge difference in cost. This, again, requires the design of a palletizing fixture.
Fixturing and workholding for production of a few parts at a time costs money. Think of it this way: Each part you have to make requires a fixture to be made for every setup. In other words, if you machine from the top, bottom, left and right of an aluminum block and want to design a fixture to hold ten of them at a time, you have to design one fixture for each setup or side you are machining. There are ways to reduce this by working closely with the designers and optimizing for work holding.
You can also integrate work-holding such that one pallet can hold the part in different setups. In other words, you hold four parts in setup 1, four in setup 2, four in setup 3 and four in setup 4. A pallet designed like this produces four finished parts per machining cycle. After each run you move the parts to the next setup location. So, the four parts that were being machined from the top are moved to the right one position and clamped onto an area of the fixture designed to hold them for bottom machining. The other parts are similarly moved one position for the next setup and the parts on the last setup slots come out of the machine for finishing (clean, debur, inspect, post-process).
If you do the math on this the scope of the manufacturing problem starts to be revealed. If you have ten parts to machine, and each part requires, on average, machining from two sides, you need to design twenty individual holding fixtures. Some can be integrated into a unified multi-setup fixture but the design work isn't any different, in fact, it can be more expensive. If the parts require machining from four sides you need 40 fixtures.
Once you are done with mechanical design of the fixtures and setups each one of them requires CNC programming and tool selection. This is a detailed process that requires time, expertise and testing, lots of all of them. In a design like that of this leg the design, fabrication, programming and testing of fixtures can easily cost much more than what the posted cost for machining one set of parts.
In short, cost reduction for something like this requires detailed and intense engagement with the manufacturing process, which would lead to design changes necessary in order to reduce manufacturing complexity, machine cycle time, fixture count and complexity, material waste, etc.
One way they could have reduced the cost to manufacture these for other people could have been to, for example, pay for manufacturing fixtures to make four to ten of these at a time. Customer #2 would then benefit from this and possibly cut their cost by a substantial amount (25% to 50%). At the limit the cost of machined parts become asymptotic with material cost + hours on the machine + aggregated costs-per-hour for the shop (people, wear and tear, power, etc.).
It's an interesting project. If the market is large enough it might make sense for someone to pay for the design and fabrication of the tooling to offer these parts at a lower price.
At a larger scale you'd have to design the parts for a casting + machining process, which will make them far less costly.
Metal 3D printing might be another option but that's not as cheap as one might think and has its own design requirements.
Thanks for looking at it. :)
My experience thus far is using a (somewhat modified) Shapeoko 3, so personally I only have a very basic practical understanding of using fixtures.
Most of my stuff was cutting MDF, for engineering prototypes though. Haven't done much with alum or anything tougher (yet), thus the limited fixturing experience. ;)
> It's an interesting project. If the market is large enough it might make sense for someone to pay for the design and fabrication of the tooling to offer these parts at a lower price.
Yeah, similar thought. They look to have added a forum (without any posts yet):
https://opensourceleg.com/forums/forum/ankle-knee-control-sy...
Asking the above there (eg have they considered potential changes to reduce manufacturing cost?), would probably get it in front of the right people.
Any interest in pinging them? If not, I'll can do it and share the response (etc). :)
My experience thus far is using a (somewhat modified) Shapeoko 3, so personally I only have a very basic practical understanding of using fixtures.
Most of my stuff was cutting MDF, for engineering prototypes though. Haven't done much with alum or anything tougher (yet), thus the limited fixturing experience. ;)
> It's an interesting project. If the market is large enough it might make sense for someone to pay for the design and fabrication of the tooling to offer these parts at a lower price.
Yeah, similar thought. They look to have added a forum (without any posts yet):
https://opensourceleg.com/forums/forum/ankle-knee-control-sy...
Asking the above there (eg have they considered potential changes to reduce manufacturing cost?), would probably get it in front of the right people.
Any interest in pinging them? If not, I'll can do it and share the response (etc). :)
Pinged them manually via email, as their forums don't have a functional log in yet (being working on apparently).
Discussion around improving the manufacturability seem welcome: :)
Discussion around improving the manufacturability seem welcome: :)
**********************************************
We'd definitely love to have some DFM help! You're right that we weren't really
thinking too much about that when we designed it; mostly, we wanted it to be as
lightweight and high performance as possible. But it'd be great to have you or others
help on the DFM side of things, especially if you're more familiar with how exactly
machining costs are calculated, and what steps / features can be modified for
the greatest improvement in cost.
-Elliott
Elliott J Rouse PhD
Assistant Professor
Director, Neurobionics Lab
Department of Mechanical Engineering
Core Faculty, Robotics Institute
University of Michigan
**********************************************Their forum is working now:
https://opensourceleg.com/forums/topic/reducing-the-cost-of-...
https://opensourceleg.com/forums/topic/reducing-the-cost-of-...
Just posted offering to help.
Being that it sounds like a worthwhile project I can see about getting involved in some manner. I have to take a look at the models in greater detail and think about it a bit.
My general thinking at the moment is that the design might have to undergo substantial changes in order to optimize for manufacturing cost. This is also one of the reasons for which you don't jump into making production tooling right away, you have to go through a few iterations before you actually know what you want to build in quantity, even if that is a few dozen units. A smart approach would create "production safe" tooling, meaning, tooling that can adapt to a reasonable degree to design changes without requiring completely new tooling design. This process requires close communication between design and manufacturing.
I'll visit their forum as soon as I have some time to dig in.
Regarding machines like Shapeoko. These are very nice and make a lot of sense for the right type of work. We have Haas CNC equipment. Once you start to worry about tolerances, finish, speed, repeatability and other factors there is no substitute for 7,000 lbs of industry-standard machinery.
My general thinking at the moment is that the design might have to undergo substantial changes in order to optimize for manufacturing cost. This is also one of the reasons for which you don't jump into making production tooling right away, you have to go through a few iterations before you actually know what you want to build in quantity, even if that is a few dozen units. A smart approach would create "production safe" tooling, meaning, tooling that can adapt to a reasonable degree to design changes without requiring completely new tooling design. This process requires close communication between design and manufacturing.
I'll visit their forum as soon as I have some time to dig in.
Regarding machines like Shapeoko. These are very nice and make a lot of sense for the right type of work. We have Haas CNC equipment. Once you start to worry about tolerances, finish, speed, repeatability and other factors there is no substitute for 7,000 lbs of industry-standard machinery.
> Regarding machines like Shapeoko. These are very nice and make a lot of sense for the right type of work. Once you start to worry about tolerances, finish, speed, repeatability and other factors there is no substitute for 7,000 lbs of industry-standard machinery.
Oh no doubt at all. My PoV there is that a person with adequate clue can upgrade and calibrate a Shapeoko3 to be suitable for projects like this. Would it take tonnes of time and effort... sure.
Would it be a production run...? Super unlikely. :)
It's the kind of thing that - for a project like this - you'd do it for an amputee friend, relative, etc. eg doable, with the right motivation
Oh no doubt at all. My PoV there is that a person with adequate clue can upgrade and calibrate a Shapeoko3 to be suitable for projects like this. Would it take tonnes of time and effort... sure.
Would it be a production run...? Super unlikely. :)
It's the kind of thing that - for a project like this - you'd do it for an amputee friend, relative, etc. eg doable, with the right motivation
I have to admit not having any time on a Shapeoko (or any other tabletop router for that matter). I've always had access to manual or CNC industrial grade machines.
I've seen some impressive work done on Shapeoko and other small machines. While this is possible, the issues are always the same: tolerances, repeatability, finish and tooling.
Without the mass to dampen vibrations it is impossible to produce smooth and accurate cuts. It is enough that aluminum is like rubber at cutting pressures. Now we add a machine with a structure, materials and mechanics that can visibly deflect a few thousands with finger pressure and the end result is predictable.
That's not to say they are not useful. They are. Definitely. For the right kind of part and project they are fantastic. At $1,000 to $2,000 they are hard to beat.
Frankly, the right way to use something like a Shapeoko for a project like this is to design the parts for the process. In other words, knowing about the limitations of the machine, design parts that will fit these constraints in the first place.
That said, a machine like this is a million times better than a Shapeoko could ever be, no matter how much it might be modified.
https://www.ebay.com/itm/LAGUN-FTV-1-VERTICAL-KNEE-MILL-w-Al...
I owned several similar machines --all purchased used for $1,500 to about $5,000 max-- before I could afford to buy my first $100K VMC (vertical machining center). These "knee mills" as they are known, can make quality parts all day long with accuracy and excellent finish. They can make nearly anything you care to throw at them. I still have one today (sometimes it's easier to work on a part on one of these than to fire-up a large machine for a few cuts). They also fit very well in a garage workshop.
I've seen some impressive work done on Shapeoko and other small machines. While this is possible, the issues are always the same: tolerances, repeatability, finish and tooling.
Without the mass to dampen vibrations it is impossible to produce smooth and accurate cuts. It is enough that aluminum is like rubber at cutting pressures. Now we add a machine with a structure, materials and mechanics that can visibly deflect a few thousands with finger pressure and the end result is predictable.
That's not to say they are not useful. They are. Definitely. For the right kind of part and project they are fantastic. At $1,000 to $2,000 they are hard to beat.
Frankly, the right way to use something like a Shapeoko for a project like this is to design the parts for the process. In other words, knowing about the limitations of the machine, design parts that will fit these constraints in the first place.
That said, a machine like this is a million times better than a Shapeoko could ever be, no matter how much it might be modified.
https://www.ebay.com/itm/LAGUN-FTV-1-VERTICAL-KNEE-MILL-w-Al...
I owned several similar machines --all purchased used for $1,500 to about $5,000 max-- before I could afford to buy my first $100K VMC (vertical machining center). These "knee mills" as they are known, can make quality parts all day long with accuracy and excellent finish. They can make nearly anything you care to throw at them. I still have one today (sometimes it's easier to work on a part on one of these than to fire-up a large machine for a few cuts). They also fit very well in a garage workshop.
> Frankly, the right way to use something like a Shapeoko for a project like this is to design the parts for the process.
Definitely agreed. :)
> https://www.ebay.com/itm/LAGUN-FTV-1-VERTICAL-KNEE-MILL-w-Al....
Heh Heh Heh
It'd be tempting to get one of those, then convert it to CNC. But... other priorities atm.
For my stuff, the tolerances of a Shapeoko are fine. For now. :)
Definitely agreed. :)
> https://www.ebay.com/itm/LAGUN-FTV-1-VERTICAL-KNEE-MILL-w-Al....
Heh Heh Heh
It'd be tempting to get one of those, then convert it to CNC. But... other priorities atm.
For my stuff, the tolerances of a Shapeoko are fine. For now. :)
BTW, the eBay link I posted is, in fact, a real ready-to-go industrial CNC machine for $3,900. My Bridgeport, which is similar, weighs 2,200 lbs. this one is a bit heavier but still just fine for a garage shop. I’ve had 7,000 lb machines in my garage.
Seriously, I’d buy one of these any time over a router, for the money there is no comparison. Let’s put it this way: The x axis leadscrew alone is worth as much as a Shapeoko.
Seriously, I’d buy one of these any time over a router, for the money there is no comparison. Let’s put it this way: The x axis leadscrew alone is worth as much as a Shapeoko.
No worries. :)
No garage here, so limited options (thus the Shapeoko, etc). :(
Btw, are you in contact with Elliot (etc) from the Open-Source Leg place? If not, there's a response to your query, with initial contact details. :)
No garage here, so limited options (thus the Shapeoko, etc). :(
Btw, are you in contact with Elliot (etc) from the Open-Source Leg place? If not, there's a response to your query, with initial contact details. :)
Thanks for making me aware of the response. I’ll go check it out and see if I can help in any way.
Very cool. I think the more open and available this technology is, the quicker we'll be able to optimize and iterate on it. I'm hoping the price of current generation prostheses plummet over the next decade as a result.
Is there a link to the source? I can't seem to find it from their website or github.
I'm guessing this is what you are looking for:
http://dephy.com/wiki/flexsea/doku.php?id=sourcecode
Transhumanists are coming. As are bionic cyberterrorists. The future is both amazing and scary!
Hmm, in both cases I think we'd need to have developed an (open-source or proprietary) prosthesis that was superior to the limb it replaces, which seems like it's a ways off still. Then you could imagine people voluntarily replacing their old-fashioned biological limbs; right now the replacement process is a lot less voluntary, and the recipients would be excited just to achieve parity...