Differential Swerve Drive

My take on a differential swerve drive. An omnidirectional wheel with stationary motors, that can drive in any direction.
I plan to use 3 of these wheels in my next autonomous robot as an alternative to traditional omni or mecanum wheels.
The model for the 3d printed parts and a list of other items can be found at www.printables...
00:00 Intro
01:10 Assembly
03:08 Demonstration - How it works

[@alexsalchemy]

Dude, that is brilliant. I was wondering why not simply use omni-wheels, but this is such a great self-contained module.

[@wildwillyrobots]

It was more fun to design this than using omni-wheels😊

[@simplegamerz9485]

Also you have more traction compared to a Omni wheel that’s the main thing about it also you have more drive power as well I’ve made a full chassis for a differential swerve drive for FTC I haven’t posted a video mostly because I don’t have the controller for it but v1 worked pretty well and v2 fixed some small problems with it by making the motor mounts stronger and using a better wheel for more grip and it seems to work pretty well mine uses 2 wheels and it works well

[@akamesama]

@@simplegamerz9485 Yeah, having messed around with omni-wheels, they are nice for clean, level environments, but struggle with more generalized design. I think most commuter cars that are currently on the market with sideways parking work more like the mark1 design referenced in the video, since you only need to allow 180 degrees of freedom. The design in this video is awesome, but I suspect the pitfall of the revised design is the maximum torque that can be produced. That said, some redesign of the gears would solve that, plus an enclosure to prevent ingress of dirt would make a fairly useful wheel.

[@ahmadhasan3258]

one of the main problems of an omni wheels is that u can push them around on those free wheels

[@DotMode]

omni wheels or mechanum wheels?

[@imaginatorstudios4175]

Nice swerve module! Btw, a simpler coaxial swerve design would also isolate the motors from the rotation of the pods. On the other hand, differential swerve is just way cooler.

[@wildwillyrobots]

I wanted a low profile module to keep my robot low, which is the main reason for using the differential design. Thinking about it now, making a coaxial module that goes around the wheel at about the same size as the differential module would be possible - but the differential was more fun to design.

[@thebloxxer22]

Clever design! Being a former FRC Team member, the swerve drive always fascinated me.

[@Champwise-xn9bb]

This is pretty cool. Reminds me of Modern Caterpillar bulldozers. They switched from a clutch and brake system to a differential steering system similar to this so they have more power when pushing around corners.

Amazing! Looks so simple and smooth that´s incredible well thought and planned. I can´t wait to see your new robot.

[@alexandrepv]

It's a great mechanical design, well done! I wonder though, how much fine tuning will be required for the controllers. It's likely that one of the motors will be under a slightly heavier load than the other, causing the direction to drift. Close loop control on both motors may be in order, but I would love to see how this develops! Good stuff man!

[@Buzz-Tea]

I thought the same. Probably an encoder is needed to fine tune the direction

[@VeydinZain]

Awesome way to think out of the box! Can't wait to see it all come together 🤖

[@TheTotalGeek]

New subscriber here. Inspiring. I’m going to try building a very simple robot with my son this Christmas holidays to get started. I hope you post more videos of your robot projects. Very cool.

[@martins.6122]

Ingenious thinking, chapeau! And thanks for having kept it open source, patents divide humanity and kill the use of free inventing and the fun of doing so...!

[@altus1226]

For the final design, you should have a toggle-able "locking" gear-set to enable maintaining a heading without relying on the differential power input from the motors. When engaged, both motors would be exclusively powering the spinning action of the wheel.

[@Skyentific]

Great design!!!

[@markklarer2447]

Very nice. Many obivous advantages over omni wheels including being much simpler design, and probably more reliable, less expensive.

[@dmitrysozinov2391]

Really interesting usage of differential mechanism that is used in cars to ajust whells speed in a turn

[@sollancer7699]

Just found your channel and I have to say your six wheels 180 is what I envision all future cars might be design around. While its cool on a drone but I feel as a vehicle platform is way better. I also think 6 wheels is the most optimal for a well balance design to get out of most situation when compare to a 4 wheels one.

[@learntostock3487]

good job this seems different than other swerve drives I have seen

[@wildwillyrobots]

I can't take credit for the design idea, but I haven't seen that many use a differential design like this, so it will be interesting to see how it performs once I put it in a robot.

[@shubhambagri5732]

I am really excited to see your fully functional robot 🤖,🐥🐥

[@MarcStollmeyer]

Back in 2012 I worked on a design we dubbed Emperor swerve for FRC competition. The design files should still be public. Our design just like this one requires very accurate sensor feedback of the wheel orientation and speed. Our biggest issues revolved around that data becoming inaccurate. We uses an optical limit switch to identify when it was pointing straight forward, without it we’d loose orientation in minutes.

[@stevenedwards408]

Amazing concept I've never seen this idea before, if you've invented this as well as creating this prototype then you have a very bright future my friend 👍🏻Excellent

[@LuisElectronico]

This was very cool, maybe add a second gear in the opposite direction but allow it to spin freely from the wheel drive shaft, this way it would do nothing in terms of motion but it would give extra stability and the wheels could handle more heavy duty aplications, overall this is a very impressive design, I want to see how it evolves

[@egecura3465]

genius idea ! Keep uploading the following videos please.

[@wildwillyrobots]

I plan to make a video on how to control this wheel, and one about the robot with three wheels. But I'm missing some parts, so I will take a while before they are ready.

[@dfunited1]

What a beauty! I see some suggestions here for sensor feedback, but it'd be dead simple to have trim potentiometers for each motor.

[@glowytheglowbug]

woah amazing design i would go with my original idea of using a turntable with gearing for the wheel but this low centre of gravity is perfect! i might try building one out of lego

[@markklarer2447]

You may want to consider an alternate gear setup that will take a couple more gears. The way you have it setup now you need both motors to spin exactly opposite to move without turning. By adding gears you can make a mechanism move without turning with only one motor. The other motor will turn it. More complicated but maybe more ideal in some situations.

[@rcdoebbora8086]

Very nice 👍

[@ONRIPRESENCE]

In the future, for whatever reason you want to be able to twist live wires freely, you can install a mercotac. They allow you to rotate the wires continuously while delivering power signal :)

[@EMILE12345678901]

so this is basically corexy but in rotational movements instead of linear

[@wildwillyrobots]

In the sense that it has two stationary motors controlling two moveable axis's, yes.

[@hopeorigins825]

This wheel motor design is pretty smart, - If both ring gears spin in the same direction : the wheel doesnt spin but turns; - If the ring gears turn in opposite directions : the wheel spins, but doesnt turn. - And by varying the speed of the one motor you can make it turn one way or the other you can make a vehicle change directions smoothly. Just a little bit bulky design and 2 motors, but other than that, it's pretty impressive. But I want to ask : when both ring gears are spinning opposite ways and the vehicle is going strait, can an external force turn the wheel to another direction ? I'd say its probably possible since such an external force would be able to slow one of the motors, this would make it even more difficult without an angle detector to know how to wheel is oriented. But that's easily fixable, you just need to put another ring gear fixed to the inner wheel support, and gear that to an infinite rotation sensor fixed to the same piece as the motors. And ... I went ahead and imagined how I would build a vehicle with this. Sorry :p

[@wildwillyrobots]

Yes, it it possible that the wheel will turn/rotate from external forces. But actually it will not turn easily, because the gear ratio for the turn is quite high compared to the ratio of the spin. I will add encoders to both motors, and that will allow me to calculate both the spin and rotation precisely. On top of that I will add control loops with feedback, so that I will be able to control the speed/position. To reset the turn angle when I power it up, I will make an initialization procedure, where the inner base triggers an opto switch.

[@ModelLights]

@@wildwillyrobots 'add encoders to both motors' Realize, this will still lose some if the middle ring is floating Instead, you can EASILY add direct encoding to the steering ring and know your direction exactly. Look at 10 seconds, to look at the vertical stack. Middle ring has its top surface clear, below the top gear ring, but above the screws for the guide bearings. Use flat head screws for better clearance. Out from the top edge of that middle ring, make a flat edge disc, then notch it for gray type encoding. Then put say 6 opto interrupters around the edge to read the code, and you can read your direction directly off the steering ring. Similar idea to the quadrature encoder of mouse, but if you do it right you can use more sensors and get X number of bits, so 64, 128, maybe even 256 positions would be doable. Probably need to rearrange the top gear drive so the motor is above so there's easy further clearance for the encoder disc, but it should be easy overall.

[@laytonpratt]

Very cool! Reminds me of a differential 🤦‍♂️. I read the title after watching the video lol.

[@greggv8]

Make the ring gears the motors. Put magnets on the rings and coils on the frame under the bottom ring and on a 3rd ring above the top ring.

[@robertoguerra5375]

That is a great idea :D Maybe in the future it could also extend the wheel downward, adding suspension too

[@jonp8015]

Are you planning on having some kind of positional sensor to check that the wheels are pointing in the direction the robot thinks they are?

[@wildwillyrobots]

Yes, I will add encoders to both motors and a third sensor for zeroing the rotational axis.

[@southern_merican]

Just when i thought u couldn't reinvent the wheel,....................here we are.

[@RichardBronosky]

I wonder if this could be used to make balancing robots. Type 1: the robot balances on this wheel that touches the ground. Type 2: the robot balances stationary on a stick with this mechanism elevated to act as a fly wheel Type 3: the robot balances in motion on a ball with this mechanism elevated to act as a fly wheel

[@BLACKM3SA]

Closer and closer to I, Robot

[@ChrisCameronPhoto]

Brilliant design. However I think a simpler answer to your problem would have been to use a slip ring to pass the drive motor wires through the steering axis.

[@InservioLetum]

Sheer genius. You'll want to patent this. QUICKLY.

[@maxxtheturtle5282]

Not to be that guy but , he wasn't the one who came up with this kind of drive mecanism. there are other version I have seen previously especially in the FTC robotics community. He did do a good job with the Cad but it's not his original idea. - kzbin.info/www/bejne/l3a3koWZobl3etE - kzbin.info/www/bejne/aWPEgYGiZb6Uq7s - kzbin.info/www/bejne/Y2jCiKmkaqapZpY

[@InservioLetum]

@@maxxtheturtle5282 Oh by all means, I always welcome correction, and much prefer it to being left in ignorance to repeat my mistake. Thank you. ;)

[@inventor121]

just one correction. it should be "Axle" which describes a type of shaft not "Axis" which is used to describe reference geometry.

[@rockyrivermushrooms529]

very clever engineering. great job!

[@satreaaji]

The hardest things is to make the twin motor to keep run on same speed, which can be solved through sensor and dimmer but the more wheels it will be more complex

[@ddegn]

Quadrature encoders would let a microcontroller accurately control the speed. I made a Mecanum wheeled robot which didn't work well until I added quadrature encoders to each of the four motors.

[@diederickb1806]

dont u need some kinda op position sensor for this? or the wheel might creep out of position

[@ethanv3653]

Very cool design! As far as your wire binding limitation on your original robot design, I was going to suggest you use brushes in the steering axis to bring power down to the drive motor, this would allow infinite rotation of the steering, ( similar to how an excavator can spin in circles indefinitely because they route the drive oil for the hydraulic drive Motors through a slewing bearing).

[@wildwillyrobots]

The problem is that in addition to the 2 power wires for the motor I also have 4 wires for an encoder. Bringing those down to the motor in a way that doesn't add noise to the encoders is not easy to do in a cheap way.

[@DesignCell]

@@wildwillyrobots This just hit my recommended and it's really cool! Came to the comments wondering about asserting positioning. You mention encoders; assuming one one each drive motor. Does this maintain precision or did you observe drift? My mind jumped to having a magnetic position sensor straddling the wheel for angle position then letting some other positioning sensor handle in environment position assert.

[@wildwillyrobots]

@@DesignCell Haven't implemented the controller yet, but from the the motor encoder values it is possible calculate the absolute position for both the rotational and spinning axis, so I don't expect problems with drift.

[@AthenaAutocross]

@@wildwillyrobots Couldn't you just do a stepped multi-level slip ring. That would put space between each of the 6 lines and still allow infinite turning.

[@kreinova2747]

Hi, i have a question. What is the maximum payload for this Swerve drive sustem if there is 3 of it. I would like to know as i want to implement 4 of it in my project. Thanksee ❤

[@wildwillyrobots]

I'm not sure. The N20 motors I used in this demo is not strong enough for heavier load. In this three wheeled design kzbin.info/www/bejne/an-Zd5mfgbFnZ9k I used stronger motors. I think it can handle several kgs before the load on the plastic parts is too much.

[@ronarmstrong835]

I think you solved the problem of having a drive for sphere wheels.

[@pauloinventostube7533]

Ótimo trabalho 👏 sucesso aí irmão 🤜🤛

[@massacrestarts1673]

Speechless 🤯

[@ahmadnuurrohim8845]

Amazing design

[@MoralRichMedia]

Very nice design.

[@Tims_Projects]

A very nice design, I love it. I foresee an issue using DC motors, if the two motors are not exactly balanced (RPM) you will have creep in steering when going straight. Do you intend fitting some sort of an encoder on the centre part for positional feedback? So that you can make adjustments to each motor.

[@wildwillyrobots]

I will add encoders to both motors. From those I can calculate the position of both the rotational and spinning axis.

[@Tims_Projects]

@@wildwillyrobots Fantastic. I look forward to seeing more.

[@learntostock3487]

I want to do this for my combat robot would this be able to withstand bumps? I would use two wheels in the middle of the bot so safe from weapons just worried about bumps and bouncing

[@wildwillyrobots]

The design itself should be fine with bumps, but obviously the plastic parts are not very strong. If you search for "frc differential swerve" there are some examples of differential swerves made from metal.

[@siggyincr7447]

It's definitely a interesting novelty. But in terms of practicality, the design used in the 6-wheeled robot is far better. The rotational limitations would be easy enough to deal with using slip ring contacts. The need for the power to transfer through 2 gear interfaces for any motion is cool, but impractical. Great for a KZbin video though.

[@MarinusMakesStuff]

Reminds me of how CoreXY systems work :)

[@kentonian]

I think you’ll need some thrust bearings for vertical loads. Another option would have been for power to be transmitted to the motor via sliding contacts at the axis.

[@wildwillyrobots]

Good point about the vertical load. My robot will be in the size of a small robot vacuum cleaner, so I think the current design will be able to handle the load, though.

[@fireworm91]

Isn't exactly how hiunday robotics designed their wheel attachments as you can see in CES 2022?

[@i-muts]

what kind of slotted bearings did you use for the ring gears

[@wildwillyrobots]

They are U624ZZ. You can find a part list here www.printables.com/model/300105-differential-swerve-drive

[@jbrownson]

Brilliant, thanks for the video

[@RichardBronosky]

Your move, James Bruton. ☮️❤️🌈

[@lelouch0reki392]

A slip ring might help with the wires limiting rotation

[@wildwillyrobots]

Others have made the same suggestion, but they are a bit expensive. Also, I'm unsure if they can deliver a noise-free signal for the encoders. Regardless, I simply wanted to try something different.

[@Rezin_8]

Epic build

[@GordonChil]

Is that an electric screwdriver at 2:16? What brand is it/where can I find one?

[@wildwillyrobots]

Yes, it is the wowstick 1F+. I use it more in this video kzbin.info/www/bejne/mInHpWueo8xlja8

[@ScottSavageTechnoScavenger]

Good stuff! Keep it coming.

[@thaerosthedragon1930]

If you get a measurement of your torque that the wheel gets I'd love to know!

[@marwanbeaino5377]

Well done !

[@beilkster]

Making the wheel a ball instead of a tire would look pretty neat

[@MaPhy]

Bravo, bravissimo!

[@bulletproofpepper2]

Thanks for sharing.

[@bio.s2903]

I was wondering, how did you design the reverse bevel gear? Or if there is a better name for it?

[@wildwillyrobots]

The bevel gear was modeled in Onshape using a feature script called "Bevel Gear Pair". It was then combined with a normal gear made using the "Spur Gear" feature script.

[@bio.s2903]

@@wildwillyrobots Could you possibly make a CAD tutorial on making one? Also if you know is there a way to do this in fusion360?

[@wildwillyrobots]

​@@bio.s2903In Onshape, once you have selected the feature script, you get a dialog, where you put in the dimensions and the gear ratio, and then the gear is automatically generated. I have never used Fusion360, but assume it can run custom scripts to generate complex parts, just like Onshape.

[@TheAnonymmynona]

The cable routing might be easier if both motor were atacht at the same side of the outer ring, but a realy cool desing anyway.

[@wildwillyrobots]

Yes. I just mirrored the part, once I had designed one of the motor mounts. For my robot the motors will be mounted slightly differently to save space.

[@learntostock3487]

what happened to the cad file it does not work when I click on the printables link

[@wildwillyrobots]

I think Printables are having some issues at the moment. It works, but download is extremely slow. According to a banner on the site, it is currently working in limited mode.

[@TheRealBustinOne]

There's a connector used in rc tanks that allows 360° spin with wires

[@benjamin_f_gates]

A slip ring? Those work, but become complex when you factor in both power and signal wires for a motor and controller. Both coaxial and differential swerve drive modules transmit rotational power mechanically.

[@brianlee5712]

You can always explore a coaxial solution!

[@wildwillyrobots]

Yes, that is on my todo list😊

[@drjeco]

I wonder what will happen when the wheel is forced to stall in either turning or motion. I suspect this designs major downfall will be unintended direction changes and wheel motion when there is stalling or a mechanical failure of a gear or bearing.. Neat design, but I don't think it will fail very gracefully.

[@cookiecai9855]

Outstanding design! I am a boop. But what makes me curious is that compared to omni wheel, what are the advantages and disadvantages?🤣

[@wildwillyrobots]

Omni wheels with rollers tend to slip on the surface. Also, you can push them around because the rollers are not actively controlled. This makes it hard to do precise navigation. This design is better suited for navigation, but compared to an omni wheel, it is much more complex both in hardware and in the software needed to control it.

[@cookiecai9855]

Thank you very much!

[@chrisBruner]

cool idea

[@OddJobEntertainment]

Really cool idea. While horizontal and vertical rotation are both easily used separately, I wonder how much use really comes from the range between. It seems like the motion gets a bit difficult to predict.

[@ddegn]

If quadrature encoders where used on both motors, the motion should be very predictable.

[@HotNoob]

this is different. buht... is it better than a less complicated rotary connector?

[@wildwillyrobots]

No, but it is more fun to create this one😀

[@FobosLee]

What’s the catch? Sure there are some limitations of design. Like loss in torque for example?

[@wildwillyrobots]

The large space required around the wheel is one obvious drawback. Another the fact that if you run the motors at full speed to drive (spin) as fast that you can, you need to slow down in order to rotate.

[@FliBaleon]

I think that I can write down the math to decouple the rotations and create a preprocessor block that given the speeds on each axis generates the speed references for the two motors

[@FliBaleon]

Actually I think I've done it

[@wildwillyrobots]

@@FliBaleon Yeah, it's not that complicated. For one motor is is the sum of the wanted rotational speeds (each divided by the ratio of the gear for the rotational axis). For the other motor it is the difference of said values. Or at least, that was what I got😀

[@babayaga8419]

How about adding a suspension

[@shhope8060]

I would like to know what size wheels you use.

[@wildwillyrobots]

Check out the description at Printables: www.printables.com/model/300105-differential-swerve-drive

[@Kinetic_CGI]

Nice design, I always wondered if there was a better way for a 3 wheeled robot. Also + 1 subs from me - lets get you over that 1000 subscriber threshold!

[@telewubby1961]

very clever designing, but an electrical slip ring and a 360 servo would have worked too

[@jaypetz]

Question is how much weight can this handle o.o

[@sammoore1979]

When making your robot, make sure to use a gyro so that the robot can be driver oriented, and not robot oriented. It makes it much easier to drive.

[@wildwillyrobots]

I will mostly be running it autonomously, so remote controlling is not an issue. But I will most likely add an IMU to improve the accuracy in autonomous mode.

[@sammoore1979]

@@wildwillyrobots Sounds good. If I can be of any help programming it, let me know. This project looks very interesting, and I am excited to see where it goes. Keep up the good work.

[@wildwillyrobots]

@@sammoore1979 I mostly got it working, as you can see in this follow-up video: kzbin.info/www/bejne/aYiToWRsq9iap6M Will make a more detailed video about how it is controlled later.

[@yogaaditya2797]

Hey, what do you think for body of robot, and gears, better use 3d print filament or resin ?

[@wildwillyrobots]

I only have experience with filament based printning. For my use it works well.

[@gstewart86]

Can't wait to see Tom Cruise race around Bangladesh in one of these puppies in Mission Impossible 17

[@szabolcs__]

I like it, but, with an out plastic gears, Intrestinf version would be, with magnets instead of cogs.

[@szabolcs__]

Sorry auto correcter

[@derekgalbraith432]

Love the design! I hope it works on a larger scale (I'm converting the model for FRC). Do you have any suggestions on how to model the big gear?

[@wildwillyrobots]

I used a gear generator script in my cad (onshape) to create the gears, and then merged a bevel gear with a spur gear.

[@Karlemilstorm]

Fedt lavet!

[@Dog-Girl-Defect]

Ok, check it you build a giant ball with a bunch of those around an inner layer of ball, you make it so you have a gyro controlled seat in the center. Boom worlds biggest bowling ball.

[@superfly8950]

That could be possibly not the greatest thing for a wheel to be overcomplicated like that. However you may have just discovered a new type of gyroscopic system? Don't quote me on that 👍

[@jadensilveira2188]

SPEEED RACER

[@JanpanTheInkling]

This could make parallel parking so much easier

[@Syzygy2048]

It's a good design, I'm just wondering if you've considered if simple slip rings would work for your original design.

[@wildwillyrobots]

Yes, but they are a bit expensive when you need 6 of them. Also, I'm not sure if the cheap ones are able to transfer the encoder signals without adding noise.

[@olw4196]

That is soooo cool

[@CreativeEngineering_]

Works like a Core XY 3d printer

[@user-yc5nm1pq4g]

🔥🔥🔥🔥🔥

[@jacobdavidcunningham1440]

well made

[@jamesgjt]

i think the only drawback is that the torque is reduce when turning.

[@brandonsilvay4822]

Another way you could use your original design is if you ran your wires from the motor for the wheel itself and for the turning up and then you with 3D printing make a column and actually put copper rings on it just like an electric motor a brush motor and then you take some old brushes or you can make some brushes using copper and Springs that will rub against the Rings in that way I actually allow you to spend 360° as many times as you want and still get the power to it as pretty much the same way how some heavy equipment works that's the same way how airplane and helicopter de-icing systems that need to use high voltage to heat up the blades work your steering wheel in your car or truck uses the same thing for the horn all the controls on your steering wheel and for the airbag it's called a clock spring for the airbag and everything

[@Duraltia]

Kinda surprised that the two small Motors have an identical or almost identical enough RPM for the wheel not to drift. Will you just run with that assumption, or do you plan to account for that with either a *_Hardware_* ( my best guess being the use of a 2 Motor single Gearbox Differential Setup driving the two large Gears with one Motor only doing the Driving while the other does the Steering ) or *_Software_* solution ( like reading out an encoder being installed somewhere )? The reason why I'm asking is because of my experience with my 1/16 scale RC Tanks that essentially have the same problem with their pairs of DC Motors powering the Tracks... My models with two independent Gearboxes always have some slight drift in them because there seems to always be one motor, gearbox or controller that is weaker than the other and even if I account for that with trim while driving forward it'll still be visible when driving backwards whereas the model with two motors and a combined gearbox will always drive perfectly straight as only one of the two motors is responsible for driving while the other does the steering.

[@wildwillyrobots]

I will add encoders to both motors, so that I can make a controller with a feedback loop.