How bad is Bowden cables for robotics: friction tests (part 2)

[@nkronert]

The scale modification is pure genius 👍

[@erikthompson404]

Interesting analysis of the cables.

[@quantumprojects5025]

So funny that you're just now exploring cables! I'm doing a research project next semester regarding Bowden cable driven robotics.

[@patrickfle9172]

Let me know if you need some input

[@aarondcmedia9585]

I still reckon the bigger issue you would have is work hardening, then fraying and finally snapping of cables under load. (Speaking from personal experience on the bike). That said, if you are wanting to pursue the best / lowest friction bowden setup, it might be worth your while to check out Jagwire products. They have brake cable outers made from Al instead of coiled steel like the standard brake outers you are using here. Pretty sure they are designed to not compress as much as standard outers. Jag wire also have inners that are designed to reduce friction - used in both gear changing and braking applications in cycling. HTH Appreciate all the robotic experimentation you share with us, it's great learning vicariously.

[@Skyentific]

Thank you for this advice! I will look at these cables!

[@aarondcmedia9585]

@@Skyentific I should add - they are expensive. However, they also have vibrant colours, including a fantastic blue that would fit the aesthetic of your current prototype.

[@martylawson1638]

Really, aluminum housing for steel cables? That's a bad idea if you live anywhere that gets hot or cold as the difference in thermal expansion will constantly mess with the brake/shifter adjustments.

[@aarondcmedia9585]

@@martylawson1638 my quick calculations says 100C temperature swing over 1m of Al housing gives you 2mm difference. What does your quick calculation say? Now consider that in cycling most Al housing usage is 20cm, and most temperature swings are 30C, and that 2mm difference goes down to micrometer deltas - again, just ballparking here.

[@digitalmunky]

Automatic spring tensioners... as used in large format audio/video tape machines will minimize risk of work hardening through normal use.

[@tomcypher3864]

I like more the steel cable system that you did earlier on this channel, seemed more precise

[@xaytana]

The Bowden tube variant has potential, but that potential isn't being met due to poor design choices, such as making the tube itself a rigid component at each end when ideally the tube needs to float along the cable to compensate for the length issues while bending. Having a standard coil spring at each end would fix this. The tube is a combination of guide and housing, where in a typical cable actuator the guides are pulleys and the housing is the paneling of the robot arm. Friction is another issue, but this could be fixed by lubricating the cable externally, in reference to the tube, and having a wiper on the cable beyond where it's lubricated, so that lubricant doesn't go beyond where it's needed. Combining these two, with a bellowed spring, would keep everything enclosed, and having a second wiper mirrored to the internal wiper would help with dust and fluid ingress of the system. With these implemented, the Bowden actuator comes closer to it's potential, though I'm sure there's further improvements that could be made. Though, the lubrication issue could also be solved by materials of better material lubricity, such as a PTFE inner tube and a cable that is drawn for a smoother surface, but in high cycle counts an actual lubricant is the ideal solution.

[@EDVDompteur]

@@xaytana I don't think a coil spring at each end would solve the problem. Because with such springs, the drive would first work against the springs before the mechanical load moves. So the first steps from the motor would only change the lenght of the bowden (puffered by the springs at the ends), but not move the mechanical load. . At the first glance one could think: "OK, then take it into account that the first x steps are lost. So I let the motor make some steps more!" - But the effect may depend from the temperature (I think so). That would mean, that the amount of "dummy steps" may differ when the temperature changes.

[@RubixB0y]

@@EDVDompteur I concur with your evaluation that the coil spring at the ends of the tubes would not be more effective than what has already been tried in these 2 videos. I don't see how that's functionally different from the initial spring-block design before the switch to brake cables. As an added point, I believe the lubricant issue was covered in the last video where additional lubricant didn't have an effect since the cables were pre-lubcicated.

[@xaytana]

I suggested this in the last video, you need to find a way to suspend the Bowden tube, ideally with a compression spring at one end, if not both ends. This will account for the length variation in the bending of the Bowden setup, as the additional springs at the end will float the tube along the cable, negating the current issue. Doing this will also keep the flexibility of the brake tube. Ideally a covered spring, like a bellows, so that the mechanism still keeps dust and fluid ingress properties at that point along the cable. The friction issue is another thing, with this much constant movement, you would ideally have a lubrication system, or components of better material lubricity. There's no real way to modify an existing cable, at least without the need to entirely deconstruct and reconstruct one, which would mean custom cables. Or, there is the possibility of having lubricant added outside of the tube, which could get messy, unless you add a wiping mechanism beyond where you add the lubrication system; such as wiper - lubrication system - tube, with this mirrored on the other end. These two ideas could combine nicely, wiper - lubrication - spring - tube/tube - spring - lubrication - wiper, for the most optimal setup. This kind of system would also need to be tensioned at the pulley, which wouldn't be difficult to implement, a set screw in the cable-fed hold in the pulley would be the most minimalist approach to this. The wiper and bellowed spring combination would also keep the cable fully enclosed, having an additional wiper mirroring the internal wiper would also boost dust and fluid ingress properties of the system. The key to understanding the issue is to understand that the tube is nothing more than a guide and a housing in one piece, nothing different than a typical cable actuator other than what the guide and housing is. In a typical cable actuator, the guide is a series of pulleys and the housing is the external paneling of the robot, whereas the Bowden tube combines both the guide and housing into a singular piece. Because of this, the housing and guide of the cable is flexible, and this flexibility comes with it's downsides that need to be compensated for, in the Bowden tube's instance this is increased friction, compensated by lubrication, and the change in length in the tube with higher flexibility brake tube, compensated by having springs at each end to allow for the change in needed length. Making the Bowden tube rigid is what produces the flaws in the actuator, when the tube itself needs to be able to float along the inner cable, so that the length issue is compensated for.

[@ARVash]

You could always do it like the human arm, put the joint motors one step before the joint reducing friction and minimize the amount of weight that is away from the shoulder. Elbow motors go on the "bicep" shoulder motor on the base, gripping motor on the elbow etc.

[@Skyentific]

Yes, I think this a good approach.

[@DavidWilliams-ov4kn]

Very helpful. Thanks!

[@xzendon]

I believe if you add a static line in parallel with the bowden tube, and use the length change of that static line to analog compensate for the change in actuator line length, this might end up being useful. Use the static line to move an idler, so that when the tube bends, tightening both actuator and static cable, the static cable pulls the idler down, and increases the effective length of the actuator line back to what it should be.

[@glabifrons]

@Skyentific You could reduce the compression load on the tubes by using more cable with a gear-reduced joint and have the cables make more than one wrap around the part to which they are attached. You'd need faster motors for the same speed of movement of the joint, but you'd need less torque to drive them too (smaller/lighter/cheaper motors).

[@Molb0rg]

Based on the result it looks like hydraulics is not necessarily a bad choice, when posible

[@Skyentific]

I think it is a good choice. But is it a cheap choice?

[@Morberis]

@@Skyentific I guess that depends on what parts you have available to you from salvage. Or if you have the skills to make or modify to produce the valving and parts you need.

[@x_ph1l]

@@Morberis The idea here is not to salvage and modify but to have something cheap, functional and possibly mass produced.

[@Morberis]

@@x_ph1l A valve converted with 3d printed parts to make it a servo driven proportional valve is that much different? Yes the hydraulic motor and the drive motors probably cant be 3d printed in plastic and withstand the pressures required but... They are also fairly cheap. But yes now you need encoders etc. and the cheap drive motors aren't particularly light. I wonder if some of the drive motors could be 3d printed for applications where they won't need high hydraulic pressures. But... The system complexity is definitely much higher than a brushless motor in the joint.

[@x_ph1l]

@@Morberis If by salvage & modify you mean to base hydraulic system on the off-the-shelf parts, like valves, and modify them to make them proportional, then yes, can be perfectly done. As you said the question is complexity. Also I haven't seen a single hydraulic system that wasn't covered in oil))

[@KevinToppenberg]

Two thoughts: 1) If motion measurement was performed at the joint rather than at the motor, wouldn't this overcome any inaccuracies introduced by cable path? I.e. drive the motor until the joint moves desired amount, rather than planning that a given motor rotation will effect a planned arm movement amount. 2) I'd like to see if the cable could be used in rotary fashion to deliver power to the arm (e.g. spinning the cable with motor), rather than by push-pull action.

[@mattanimation]

would be interesting for your to explore the cable driven arm design that Naver Labs has for their robotic arms. They aren't using a bowden cable but cables none the less.

[@Skyentific]

I have a video about this arm. :)

[@MisterKaen]

You rock brother. Love the videos Very well explained

[@martylawson1638]

The cables follow the Capstan equation so minimizing the total bend angle of the cable is key. The cheapest cables use a raw cable surface and nylon or polyethylene liners. If you're willing to pay more you can get cables that are drawn after winding so they have a smooth surface, and cable jackets that use a PTFE liner for lower friction. Finally, there are 1-2 makers of brake cable housing that use the "shift cable" construction. Maybe they offer the best of both worlds?

[@xaytana]

If we were to float the 'brake' tube along the cable via springs at each end, so that the overall length can extend and compress to compensate for the tube itself having a length difference, would this mechanism not find the ideal length and bend angle itself, due to the weaker of the springs accounting for length, making the tube's internal spring the harder of the two to actuate, thus a higher resistance to bending? The tube needs to float along the cable anyways, there's no reason for it to be a rigid component, and only produces flaws in the design of the Bowden actuator. Springs are the ideal solution to this, as they would keep the tube anchored at each end, yet the tube can still float along the cable. If this solution provides a secondary solution for the bend angle issue, then there's even more reason to implement it.

[@jacobdavidcunningham1440]

0:54 I think it's because of the slack right (space between cable and walls), it goes diagonally inside the tube while rolled up and taking up some of the cable length sideways

[@Skyentific]

Not really. Shift cables also have this slack, but they don’t have such problem. There is no much slack, so the effect which you mentioned is negligible (and can be absorbed by the elasticity of the cable/bowden tube.

[@Skyentific]

But this is a great point!!!

[@jacobdavidcunningham1440]

@@Skyentific I see, when you took the cable apart I didn't realize it had those "spring-like" walls inside. Still hard to imagine the black part shrinking/contracting length wise.

[@titter3648]

Just put a encoder on the joint end and then the changing length will not be a problem. And the hall effect magnetic encoder IC's is really cheap too. And for the friction i always apply a thin coating of grease to the wire for shifter cables on my bikes. And it really makes a difference. For example in my dropper post seatpost the wire would not even return with the return spring when it was dry, while when it was greased it easily returned.

[@Skyentific]

There are already grease on these cables. I tried additional lubrication, but it does not change the friction.

[@ElectricGears]

Even if the encoder is cheep, it means you need to use a more expensive close loop motor driver. You are also not solving the problem of the changing lengths which add a whole other layer of complexity to the inverse kinematic calculations needed to position the end effector accurately. It is possible to compensate for backlash in a gearing system, but these cables introduce *backlash that constantly changes based on other joint angles*. Technically it could be modeled and compensated for but it would be a nightmare of extra complexity

[@titter3648]

@@ElectricGears No you just add or subtract the position given to the motor controller based on the difference the central microcontroller sees from the feedback of the two encoders (motor and joint encoder) continuously. So it is no more complexity and no more HW other than the joint encoder. Or you could have just the joint encoder and control the motor from that, but that might mess with the PID parameters.

[@cem_kaya]

what about 3d printing Bowden tupes ? cant you put your cable in it ?

[@benwechter6488]

Great data collected!

[@Blowjin]

People who planned build robots arm with a Bowden cables: 0

[@djalmaribeirodrone]

You who are a master of robotic arms! Help me with a question. What would be the biggest motor of the arms that make movement with big cameras like for example KIRA?

[@AReed030188]

Couldn't you just apply the capstan equations to figure out or at ballpark the friction forces and actuation forces based on the wrap angle any time the wire goes around a corner?

[@Skyentific]

Great question. I think in some extent yes. But the capstan equation is for holding (static) friction force, and here I concentrated more on sliding (dynamic) friction force.

[@dubsar]

Is it possible to use graphite as a lubricant, or a sealed tube, or both?

[@stefanguiton]

Great video!

[@lemonysnicket6153]

Man, that's a LOT of hazelnuts in your shopping

[@Bit-while_going]

It reminds me of a problem I once had with a clutch cable. If the cable was bent to much, it will wear the cable down and break it. Fortunately those don't really move from their ideal positions. Maybe it you had one cable pull for half the range of movement and another one pull for the other half? And then they could work together for extra strength. The friction would be reduced and the strength increased.

[@trapiturette]

i am late to the show, but perhaps you could try Campognolo cables, the are way better than most other brands. love your channel!!!

[@sergeidmitrik922]

You could try to use polytetrafluorethylene (fluorine plastic) tube insert to significantly reduce friction.

[@roidroid]

9:36 Macadamia Nuts? It is native to my region :) There is also another nut that is similar size tho, i forget it's name.

[@satchelsieniewicz5824]

what about frequency response from input to output??

[@AlexanderSauerPlaysWithHeroJr]

love ur videos btw.... thank you...

[@patrickfle9172]

I'd use a solid bowden housing like Nokon, which is more flexible than conventional shifting cable housings and has less friction. Still, friction rises approximately following Euler-Eytelwein formula for belt tension with the accumulated bend angle. de.m.wikipedia.org/wiki/Euler-Eytelwein-Formel

[@МихаилТокарев-н8р]

Хорошее видео, классые драйвера!)

[@TheNamelessOne12357]

Why didn't you try to transfer rotation via these cables? Put reducer to the wrist, motor somewhere else, and transfer rotation from the motor to the reducer with this cable.

[@maeanderdev]

Maybe you could get away with the bowden cable length issue by using rotation sensors in the wrist with a control loop

[@thePavuk]

Know it's old, but there are shift cables with teflon coated wires so the friction is lower.

[@santosvella]

You should try Graphite spray in the tubes. They make Graphite spray for bikes.

[@cringineering]

Lubricate the cables and check friction?

[@Skyentific]

Done. The same result.

[@Blowjin]

colored hex keys is cool

[@XXCoder]

Siiigh still waiting for autocaptions. It sucks when I can't enjoy a video for days and sometimes weeks.

[@XXCoder]

Seems you removed captions from all your videos?

[@jindramares9550]

What about use just ther plastic inside

[@skinife]

what if the cables are lubricated?

[@Skyentific]

They are already lubricated. I tried to add more lubrication, - it does not change much.

[@gatoninja4387]

@@Skyentific decrease the length of the outer spring by half and repeat the tests

[@manoelvictordonascimentogu5934]

Muito bom!

[@ElectricGears]

Perhaps this has been suggested in other videos, but you might expand your evaluation into rotating cables, like the kind used in the flexible shafts for small rotary tools. I would have them drive a fine tooth worm wheel in the same place as the surfaces for the existing pulling cables. The cables in the cheep flexible tools are probably single layer so you can really only apply torsion in one direction. I believe they also make better quality ones that are double layer (similar to anti-rotation lifting cable). I'm not sure how stiff they are, but it might be an interesting experiment. It might also be possible to use two cables with opposing twists and preload them against each other. If you have a lathe it's fairly easy to make a worm wheel with a tap and use a bolt as the worm gear. kzbin.info/www/bejne/gGHSZIpqlLGFeNk If you make the worm wheel out of plastic I think you could 3D print a fixture that would all you to slowly move the plastic blank into the tap while it was driven by a cordless drill.

[@EDVDompteur]

@ElectricGears: Nice idea with the rotating core inside the cable, but I think it would cause two new problems: 1) The cable may only work in a good way for one scecific direction of rotation (if not optimised for both directions). 2) If the direction of rotation changes, the motor must first overcome the elasticity of the drilled core inside the cable before the mechanical load can react. Maybe I'm wrong, it has to be tested.

[@WillianMai]

from what you have shown, I think bowden cables are a really bad choice for precision but tensioned cables (not bowden cables) wouldn't be a bad choice. I'd rather be interested in your opinion about precision and repeatability of this one: kzbin.info/www/bejne/q6Wnkpqlqt2Aa7c instead of bowden cables, which seems pretty unusable for robotic arms.. this arm joint on that video was amazing in my opinion, I'd love to see more of that. I think there's room for an open source low cost arm using that kind of approach, and we all would love to have access to this kind of tech. I am developing an open source hardware abstraction layer for integrating robotics with computer vision and AI. But an arm is something that I cannot afford and don't have available time to design from ground zero yet.

[@aretgh]

если тесты будут продолжаться то тросики есть тормозные есть переключения есть с тефлоновым покрытием(черные обычно) есть и нержа рубашки(внешняя оболочка так называется в вело теме...) все тож разные....взять просто самую дешевую и шимано оригинальнуювелосипедеJagwire) ну а самое крутое что в руках держал это YOKOZUNA!!!!!после установки их я как

[@Skyentific]

Я раньше и не знал что их столько разных :). Я стараюсь найти дешевое решение, поэтому совсем экзотические кабели пробовать точно не буду. :)

[@aretgh]

@@Skyentific тогда пробуйте шимано средние и подороже Ну и джэгвайр Интересно будет сравнить с самыми дешевыми(тк руками на велике разница есть точно!) Ну и тросики!!!! тормозной и переключения разные по толщине и плетению !!!

[@SergeiSugaroverdoseShuykov]

@@Skyentific их много разных только принцип по сути один - стальной трос гуляющий по фторопластовой трубке, так-что без фрикционных потерь в принципе не обойтись, в теории разве-что трос жестко упаковать в фторопласт и гонять его внутри другой фторопластовой трубки т.е. трение будет не сталь-фторопласт, а фторопласт-фторопласт

[@key2010]

omg, why do you have so much chestnuts?

[@eelcohoogendoorn8044]

Very interesting experiments; but indeed mostly negative results id say. Have you ever looked into mckibben muscles? Kevlar sleeve is quite inexpensive, and it is possible to make nice zero-sealing-surface closed hydraulic actuators. Rather than using a valve manifold and all the control problems that brings with it, you can make it semi-direct drive, without any valves and just one motor per mckibben-pair. Does not make the actual motor components any smaller, but it should keep the arm itself very lightweight and allow you to place the motors almost arbitrarily far away.

[@thiagokeizo]

Hi!

[@Skyentific]

Hi! :)

[@genephipps6421]

I am still not 100% clear why you are using cables intended for a push / pull system when you are set up as a pull / pull system? Because you are pull / pull you do not need all of the cable sleeve--just smaller fixed sections of it in key locations. It seems like that might eliminate all of your problems.

[@PiefacePete46]

Gene Phipps: I have been pondering what you have said; I am familiar with bicycle shift systems that use exposed cable on the straight runs, and short lengths of sleeve where the direction changes. In the instance I am thinking of, the handlebar end has a lever with a detent mechanism, while the derailleur end has a spring. The cable is always in tension. I cannot think of a mechanism that puts the inner cable in compression, as would be the case in a push / pull system. Judging by the results of the tests shown here, the friction losses are greatest at the curved sections... I suspect the straight runs contribute little to the total losses.

[@genephipps6421]

@@PiefacePete46 a continuous sleeve allows the cable to operate as a push pull system--as long as there is no place for the cable to deflect outward it works in either direction. In a pull / pull setup, as long as there is something to redirect the force at the bends--a short length of cable sleeve or a pulley--the system will work fine. Remote steering for outboard boat motors is an example of push pull as it uses a single cable. Light aircraft typically use a pull pull system. I guess the thing I am really unclear on is why he would have tried a *pair* of push pull continuous sleeve cables in the first place as the results are pretty predictable. The geometry for constant length cables is also not difficult to correct for. As long as the pulleys that redirect cable direction are centered on the arm joint axis cable length remains constant. I am not being critical, more thinking out loud as I find mechanical puzzles like these quite interesting.

[@gatoninja4387]

makes a 3 video with state-of-the-art lubricants and the system can also be improved so far it is the best system for price, volume, duration, strength, and low weight,

[@gatoninja4387]

strength test now

[@gatoninja4387]

It is the cheapest and easiest system to do

[@ReginaldCarey]

You failed to include a control in your experiments. Also, shouldn’t your use of cables mimic tendons? You are relying on stiffness and incompressibility of the cable sheath. That approach does not mimic nature.

[@tmpEngine]

why not just use the cable without the housing, no friction, profit

[@tmpEngine]

my mountain bike cable is almost 70% with no housing, just goes through the frame naked

[@AlexanderSauerPlaysWithHeroJr]

you apply BOWDEN TUBES WRONG , ...you used tube/ as flexible arm.concept... the bowden tube outside stoppersclams should be connected on either end of the SOLID ARMATURE ..the rotation.gears connected to cable ends shoud "sit' on eiter end of the rigid arm.... the outer tube isnt supossed to move at all.... be part of arm ..rotating a joint on eiter side or next arm..(own dedicated double cables) should only affect the position of the cable ends cause gear cable input is right next to arm/end bowden tube ...

[@VideoStefan17]

a comment! :)

[@VideoStefan17]

@@Fleurlean4 uh, a interaction! ;)

[@sostoytelnost]

Русский чтоли?

[@nou4898]

third comment