Experiments with Cycloidal Drives

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
the current version of my robot dog open dog version 2 uses belt reductions from the motors to the leg joints belt drives are very easy to build because all the parts can be 3d printed the belt grips on multiple teeth and therefore spreads the load across multiple parts of the print provided the belt can be tensioned correctly with idlers so it doesn't slip this makes a reliable drivetrain although i did have some issues with the pulleys on the knee motor breaking i reduced the size of the bearing in the middle so now there's more plastic around it and now they're holding up this bearing helps support the end of the pulley so now the motor mount can't bend under the load of the belt and it's made everything operate really reliably however i only have five to one belt reductions which means there isn't an ideal amount of torque and the door can't carry much load before its legs collapse one option is to increase the motor size although because the motor will be bigger this means adding more mass and also bigger batteries since the battery i've been using is right on the edge of the currently can source the other option is to make a higher belt or gear reduction ideally increasing the ratio to ten to one but this would mean that using belts the big pulley would need to be twice the diameter either that or we'd need a two-stage belt reduction which would take more physical space and need to be tensioned correctly lots of robot dogs such as mit cheetah use planetary gearboxes which allow a higher reduction in a smaller space and talk to be conveyed more efficiently than a simple spur gear however i'd like to be able to 3d print everything and this also makes the designs more accessible to anyone else wishing to reproduce them open dog's legs move quite quickly and have to support its entire mass when the feet hit the ground so i'm not convinced that a compact 3d printed gear will hold up especially the one in the center for planetary gearbox attached directly to the motor but there's another option which several viewers have suggested in the past the cycloidal drive this uses a cycloidal disc spinning off center as a kind of single gear reduction this could be made to be quite heavy duty and so a 3d print should hold up to the torque required i've seen various projects that use cycloidal drives and although there are a few youtube videos out there about walking robots with these types of reducers i'm having trouble finding one where there's actually a video of it all up and walking so i have several questions that will need answering how quickly will it wear out i've seen some destruction testing videos so i know it'll happen eventually but how can we make it more reliable and also can it be back driven this is important to give natural spring in the leg of the robot using motor holding power to hold the joint in place and having the motor increase the amount of power to the motor to spring it back into position this is one of the fundamental reasons that open dog version 2 works at all because the belt drive is easy to back drive so let's talk about how cycloidal drives work and try to build one that's reliable let's imagine a ring gear with some teeth around the outside and a smaller gear with some teeth on the inside as i rotate the small gear around and around the outside it turns around in this case many times the ratio of this rotation is the outer teeth minus the inner teeth divided by the inner teeth which in our case is 15 minus six all divided by six which is 1.5 i've now marked one tooth on the small gear so you can see how many times it rotates and you should see that each time i move it around the outer gear it rotates one and a half times which is the answer we got from the formula however this means it's going faster so we gearing up and actually what i want to do is make a ten to one reduction however to make a ten to one reduction we need eleven teeth on the outside and ten teeth on the inside but that won't rotate without the gears binding so we need a slightly different solution so i made a new prototype with rounder shallower teeth that can pass each other more easily this one has 11 teeth on the outside and 10 on the inside and now you'll notice that every 10 times i rotate the inner gear around the outer gear it actually does one complete rotation giving us a gear reduction of ten to one however it does sound a lot like grinding plastic which is probably why 3d printed cycloidal gears wear out the best solution to this seem to be to fit bearings all the way around the outside so that the plastic disc can roll around them so this seems to run pretty well and i'm pretty happy with the action that we've got there so far if you look closely at the bearings you can actually see them rotating that means they're doing their job as the cycloidal disc rolls over them at the moment though my bearings are mounted on m4 bolts which means they're slightly loose ideally these need upgrading to toleranced m4 steel bar so now we need to drive an output from this and this is driven by the purple pegs you can see in the animation however i want to put bearings on those as well which means the holes need to be a certain size and the size of that hole is the diameter of the off-center rotation of the cycloidal disc which is nine millimeters plus the peg which in our case is a 12 millimeter bearing giving a 21 millimeter requirement for the hole however with those holes cut in the disc we have hardly any disc left and this part looks pretty weak and it will probably break after not much time thanks to 3d fuel for the filament for this project and lots of other projects so check out my channel for more 3d printing projects and check out 3dfuel.com so it's time for a larger version which means the cycloidal disc can be ever so slightly larger this one has a recess for a bearing in the bottom to mount the output shaft and you can see that i've got my bearings mounted on pegs which will go in the holes in the cycloidal disc having moved the other 11 bearings over and made a cycloidal disc that looks a bit more substantial we can see that the mechanism works quite well giving us uniform rotation on the output shaft as we rotate the cycloidal disc it also looks very much like it's going to be back drivable but we won't really know until we attach a motor and because the disc is quite light hopefully there won't be too many vibrations as it spins off center when the dog legs move really quick it's probably time to add a motor so i'm using the turning g multi-star aerodrive 9225 90 kv motor this is a spare one from open dog version 2 so it's exactly the same as the existing robot that motor obviously spins at the top and the base is stationary so this is bolted into a kind of cup and that allows the top to spin freely i've also left a place for the cables to exit which come out of the bottom of this motor and of course fitted to that is a bearing which runs on an offset cam so that's going to move the cyclodal disc around on its off-center rotation in a nine millimeter diameter i build spacers into the 3d print so that the bearings all sit at the right height however on the outer bearings i really need to increase that by a slight amount so i'm using some nylon washers even with the bearings at the same height if i back drive the mechanism the cycloidal disc does pop out so it'd be better if we had two bearings in the center one above and one below constraining it just like a skateboard wheel it does however rest on two surfaces above and below so it'll be fine when everything is assembled and that's nothing that some silicone grease can't fix i'm just applying a bit of that to the surface above and below and that should be fine for testing now it's time for a quick ad from the video sponsor which is phantom wallet phantom wallet is available in a number of sizes and finishes there are two lineups phantom r and phantom s phantom wallets include a special mechanism that fans out your cards for easy access there are lots of accessories for customization available which attach to phantom wallets such as a coin holder money clip key holder a multi-purpose silicon band id holder and tracker phantom wallets are very stylish and also compact so they reduce pocket bulge considerably check out phantom wallet at the link below in the video's description box and use my coupon code bruton for 10 [Music] off i've made a top for the mechanism and that locks into the 11 bolts for the 11 bearings around the outside of the cycloidal drive it's a pretty tight fit and it locks down nice and tight and also stops the cycloidal disc from popping out for now i'm just going to put three screws in but it has actually space for 11. and we can see that that seems to run nice and well with the motor separated which can happen at any time the reduce is actually pretty compact and of course any motor can be put on there and of course the bolts that come from the bottom all the way through the five bearings through the cycloidal disk output shaft are the same ones that pop out of the top so the output shaft is basically double braced on those two bearings which should be pretty reliable so i've reassembled everything back into one unit and now we can see that it definitely is back drivable and i've left a hole in the middle here so we can see the cycloidal disc doing its work there is some backlash though and that's just down to the tolerance of the cycloidal disc so ideally all of the holes including the one in the center where the bearing is need to come in a little amount the weight of this unit is under 850 grams which is fine we'd need 12 of them for a robot dog so that's completely within scope for the total weight i'll be testing the assembly with the same good old 6 cell lipo that consoles plenty of currents however i don't have any o drive spare at the moment so i'm going to be using a skateboard esc it's one of the vescs which is the next best thing and i've got an arduino in a box with a knob attached so i can vary the servo pwm signal sent to the motor driver there's no feedback encoder or anything like that which is what we really need for positioning this accurately but will be good enough for a load test the whole assembly as you'd imagine with that off-center cycloidal disc does shake a bit and if i hold it loosely in my hand you can see it if i put it down it will walk along the mat as i power it up although some of that is down to the motor recoil as i power up and down that big brushless motor one way to get rid of the wobble would be to have a second cycloidal disc 180 degrees out of phase to counterbalance the off-center cam and the off-center disc however this would need a cam with a crankshaft in so we can spin both of those equally off-center in opposite directions so that's a bit of a different design which would come out slightly deeper so that seems to be running okay on the vesc but how much power has it got well i'm just running at a moderate speed and i've limited current to 20 amps but i can't stop that with my hands so i went on to make a bracket to attach to the front of the motor and a lever so it's easier to grab hold of and we can do some testing i screwed that to a piece of wood and clamped it to the table and we can see that's more than fast enough to act as a dog leg joint i don't think there's any issue there and i'm only running on 24 volts i could run up to 48 volts and run it twice as fast but that seems more than agile enough but how much torque does it have i decided it was a good idea to tie a four kilogram weight on the end of a string and tie that to the lever and the lever is quarter of a meter long so that doesn't sound too good let's see what's happening it appears that the disc isn't even rotating and the cam is slipping for some reason if we look at the back of the motor we can see the actual whole motor flexing on its mount well it's okay off load again and it seems to be running okay although it's making a bit of a different sound to the sound it was making before it's running quite a bit rougher and in fact what's happened is the cam has snapped off entirely which is kind of what you'd expect because it's only made of plastic and it's got that massive offset on it the rest of the mechanism seems absolutely intact though so the cycloidal disc is still fine and everything seems to be running smoothly without the motor attached so no problems there so the best improvement i can imagine making at the moment would be to have that camshaft running off the motor actually long enough to reach all the way through the gearbox and have it braced with a bearing at the other side that means that it won't skew so much and there won't be so much load on it that will cause it to snap off and the best arrangement would be more similar to the animation where the green shaft in the middle doesn't actually spin off center but it has a cam mounted on it that does and that means that drive shaft could be braced with bearings at each end and there would be no skew on it attached to the motor it also means the cycloidal disc would need to be bigger so the hole is bigger but we could mount multiple cycloidal discs on the same shaft but that's going to be the next version of the design i am going to come back for a version 2 which is going to be slightly bigger so we can make that hole in the middle of the cycloidal disc bigger we can run a straight shaft all the way through brace with another bearing at the front as well as the motor bearing at the back and we can then put a cam on that shaft or multiple cams for multiple cycloidal discs the whole thing is going to get taller as well so we can get multiple sets of bearings in to run those disks on but that should convey force more efficiently and put less load on the cam on the motor so that is going to be another project that i'm going to come back to but for now i'm going to publish this version as open source on github and that is a solid model that you can edit more easily if you want to play around with it rather than stls which are a mesh which are pretty hard to modify so if you'd like to support me through patreon or youtube channel membership those links are in the description below as well and patrons and youtube channel members can get access to all the videos up to a week early as well as sneak peeks and pictures of what's coming up alright that's all for [Music] now [Music] you
Info
Channel: James Bruton
Views: 443,073
Rating: 4.9451885 out of 5
Keywords: cycloidal drive reducer, 3d printed gears for robotics, 3d printed gearbox, 3d printing robot parts, tough 3d printed mechanical parts, experiments with cycloidal drive, strain-weave gear, adding bearings to cycloidal drive, cheap reducer for robotics, strain testing for 3D printed parts, cycloidal drive with bearings in
Id: pWMB5VbLb6w
Channel Id: undefined
Length: 15min 5sec (905 seconds)
Published: Mon Mar 15 2021
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.