a thoughtful web.
Good ideas and conversation. No ads, no tracking.   Login or Take a Tour!
comment by briandmyers

The leg-joints solution you've described is pretty-much exactly how it's done in the Rex exoskeleton (I worked there a few years back).

The actuator motors push and pull like muscles do - 10 in all, 2x2 for upper leg, 2x2 for lower leg, and another pair near the hip for lateral motion. The motors were made by the same crowd that made the motors for the Mars rovers. Hydraulic power is out, due to risk of hot fluid / patient proximity, so it's all-electric. No actuator wires though (motors were directly connected to the skeleton), there were enough problems with system-slack as it was; wires would make the problem worse (it's critical to know physical leg position pretty exactly).





Killerhurtz  ·  3180 days ago  ·  link  ·  

Not saying actuator wires, though. I'm saying wires attached to actuators - the actuators pull the wires, and instead of pushing back, to do the opposite movement you pull the opposite. Except that I can see actuators being a bit harder to use, so I was thinking more along the lines of a stepper motor assembly driving a bar (not unlike bicycle brakes) or a pulley (not unlike securing cord on flatbeds) that would pull the steel wire.

Though I'm glad I actually did figure out something that actual, certified engineers have been using. Makes me feel clever.

briandmyers  ·  3180 days ago  ·  link  ·  

I got what you meant; maybe I wasn't clear. The problem is that wires stretch. These are stepper motors.

Killerhurtz  ·  3180 days ago  ·  link  ·  

Now I have a question. Just how much weight/force is involved into that suit for steel wire/cable stretch to be a considerable issue?

briandmyers  ·  3180 days ago  ·  link  ·  

I'm a software guy so this isn't my area of expertise; I can't give you a numbers answer. The motors are quite powerful though, they can easily break bones, so user-safety was an important concern (especially since most users of the devices had no feeling and wouldn't know they'd been injured right away). However, in general terms - you command the motors to go to a particular position and stop; in zero-G, that would be the end of the story, but under gravity, there will be "sag" or "slack" in the system - it moves a little bit. It takes very little difference between "where I am" and "where I believe I am", to affect the machine's balance. Every linkage adds a bit more "looseness". As far as I know, wires were never even considered, because it's a pretty high-spec'ed wire that doesn't stretch at all. Hope that makes sense.

Killerhurtz  ·  3180 days ago  ·  link  ·  

That can't possibly right, though. No matter where I look, the math doesn't support it - using 1x7 galvanized, according to this page, and assuming an extreme load of 600 pounds, it would still only stretch by ~0.01% of the cable's length - which, for a cable of let's say 18 inches for measurement, has about the same stretch as it has diameter - a quarter inch. For stretch-proofed cable, this calculator (assuming eighth of an inch diameter, 18 inch long 1x7 cable and generous 120 pounds-force load) gives an even drastically insignificant stretch - 4 millimetres.

And unless they used slow, screw-gear actuators, I'm pretty sure that the inaccuracy is in similar numbers.

Larso  ·  3179 days ago  ·  link  ·  

I am a Mech E. While that amount of displacement may not be noticeable to the user, it is noticeable to the control system and could certainly throw things off, depending on the control system used. It would need to be accounted for.

Killerhurtz  ·  3179 days ago  ·  link  ·  

It would need to be accounted for, sure. I'm not saying it's not. But it's accountable for. And hence why I say it's not an issue. Hell, the control systems might not even have to deal with it if you design your thing so that the cable stretch is beyond negligible.

Then again, I realized we were talking about two completely different systems. He was talking about exosuits to practically replace people's mobility. I was more thinking of strength amplification where the suit makes your movement stronger - but the accuracy is still governed by the human.

briandmyers  ·  3180 days ago  ·  link  ·  

I'm not a mech, but 4 mm is far from insignificant. Also, you'd probably be surprised at how much force goes through the hip joints - I don't know the numbers, but that hip joint really had to be made tough. We destroyed a number of them.

Killerhurtz  ·  3179 days ago  ·  link  ·  

4mm out of 460mm is not insignificant? It's less than 1%. Most people I know would call that number a rounding error.

As for how much force goes through the hip joints - I am aware, that is why I was suggesting to apply the force away from the joints. Do you not know of the lever principle?

kleinbl00  ·  3179 days ago  ·  link  ·  

This is going to get you into trouble: you're talking to someone who was actually involved in systems very similar to your ideas, and is giving you boots-on-ground examples of why the things you haven't considered are preventing your ideas from becoming reality. Your response is to argue with his reality.

I am a mech so I'm going to give you some homework.

Take a look at your hand. In particular, take a look at your index finger. Point it, then make a fist. Take note of where the extensor tendon anchors to your wrist. Point your finger and measure from your wrist to your second knuckle. Now make a fist and measure the same distance.

What percentage difference are we talking about? I measure about 4 1/2" and about 5". That's a difference of about ten percent... and that difference governs 90 degrees of flexion of the knuckle joint. 1% stretch, then equals a loss of ten degrees of precision in positioning your finger - that's parkinson's territory. And that's just a finger. Imagine trying to run if you could only plant a leg within about 10% precision.

Beyond that, motors don't lend themselves well to linear torque. As demonstrated, your total flexion for a rotational motion of 90 degrees is about ten percent... and the distance extended is handled much better by a linear ram (ANY linear ram). Motors are designed to spin, not rotate through a limited arc. And if you're actually moving your cables enough that they're being spooled, they've just become braided or twisted and are undergoing substantial deformation, which means their failure rate just skyrocketed.

briandmyers didn't have to tell you this. He's seen it and knows there are problems. He's worked on this, and he's a clever dude. You, on the other hand, chose to dive into sarcasm in a comment thread where you disavow MATH for fuck's sake. You can choose not to believe in math but I'm here to tell you, friend, math believes in you... and when you whip out "the lever principle" in a discussion of mechanical engineering in which you don't know the players, your interests are best served by not being a smartass lest someone comes along to demonstrate your dearth of knowledge.

Killerhurtz  ·  3179 days ago  ·  link  ·  

It's not going to land me in trouble for a few reasons: first, I don't particularly plan to implement those in a lifetime. Second, and it's something I apologize for, there's a point I overlooked: I realized that we were talking about two completely different applications. He was talking about replacing an user's mobility in medical fields. I was talking about augmenting an user's already existing mobility - this idea was mainly to amplify the strength of an user that already had full mobility.

As for that stretch - that's 10 degrees only if you use a very basic driver. The wonderful thing about electronics is that you can account for that stretch. Or you could use materials science and use, as I mentioned, pre-stressed cable that does not stretch as much, or even design your system so that the cable is already under a tension high enough that any further stretch is even less of an issue (though I do acknowledge that this is not a fail-proof execution either).

And as far as your example go - I also have issues with it. The one percent stretch I mentioned was in the case of an extreme load on a thick cable at a very high weight - something that would be found only in very overweight people or military applications. The aforementioned finger would, in all likeliness, never even reach 0.5% of that load. Using that same linked calculator, with a 6-inch 0.11 inch cable under 30 pounds-force of load, it's 0.02% of cable length - an accuracy of about 0.018 degrees, which I do believe is on par, if not finer than human motor control.

For the motor issue, I'll give you that - but with a stepper motor (or an array of stepper motor) you could have enough control to make it useful. And I used the term spool to refer to a circular thing that turns and affects the "length" of a cable - or, as I mentioned, a bar.

As for the "lever principle" thing, it was not sarcasm. I have honestly seen people with higher education, like software engineer, that forget about such basic principles. And I did not disavow math - math is the only thing that matters in engineering - I just consider all the options, like actually designing things to avoid the bigger math problems like stretching. But yeah - the lever principle comment was not sarcasm - it was merely to ensure that we had a similar knowledge base. It's not because he's clever that he didn't forget about things. Sorry if I came across as a smartass.