With my teams season now over, myself and a few others on my team have started playing around with the idea of making a swerve drive over our offseason. Here is my first iteration. It is inspired by the work that 2079 did back in lunacy. The part I love most about this style of swerve is not having to design in and mount thrust bearings as the plates of HDPE or similar plastic takes its place. After reading some of the points 2079 made in this post, There are a few design changes I will be making. The main one being making an inlay around the grooves that the HDPE rides on to aid in reduced friction while turning the module.
For machining this we have a new CNC that should be arriving relatively soon equipped with a 4th axis allowing the indexing of this part (such as the cog pattern near the top) to be done in less time and be done more accurately than with a conventional mill and an indexing head.
Any comments/suggestions greatly welcome as this will be our first time dealing with swerve the inherent challenges it brings with it.
I have done some research into swerves (but not too much), and this design looks pretty nice. The only criticism I have to offer is that it looks really heavy. The lightest swerve modules that I’ve seen have minimal framework around the wheel, and that module just looks huge and heavy.
A swerve drive is an inherently heavy system to put on a robot, and if you plan to use these on a robot in upcoming seasons then you will want to reduce as much as possible. I recommend checking out the swerves of 1640 and 16, as these are the lightest drives that I know of.
According to solidworks (of which I’ve found to usually be a little light) the shell of this module is 1.3lbs. At the moment weight isn’t a huge concern. This will be the type of project that would need to be iterated to find what works best for our purposes.
The main reason behind the design/amount of work to machine this is because many of us on the team are excited about this new 4th axis. I have been wanting try swerve for a while now and this was the perfect opportunity to try out both things. As for the height, it was an arbitrary dimension I used and can easily (probably will) be changed soon. Shortening it should also help in reducing some weight.
This looks pretty cool. It’s sort of similar to 696’s swerve, which is one of the coolest I’ve seen in a long time. Their bearing setup works very nicely, but they had to do an incredible amount of machining to make the module frames.
A few suggestions/ideas:
If you’re trying to save weight, you could likely make the whole thing shorter and use gears where you have chains. Vex sells gears with the bolt pattern in them, so you can still have your dead axle.
I’m a little worried about the two plates that drop down and hold the bearings for the bevel gears. The .375" ID flanged round bearings are typically .220" thick or so IIRC, meaning the two fasteners are likely 6’s or 8’s. Those plates look like they might have trouble support the bearings hanging below them, but it’s hard to know from just looking at a few screenshots.
The vertical shaft has only one bearing. This screen shot might be missing the other, but I’d highly recommend adding another one.
The bolts holding on the top plate at the moment are meant to be 10 bolts but may need to change those plates to give it a little more meat to hold onto the bolt. As for the vertical shaft and single bearing combo, that is something I have been working on but am struggling to find a way to mount a second one in a clean, strong and efficient manner.
Also the reason I am trying to stick with the chain rather than gears is because the chain is cheaper for us. I need be we can make the sprockets ourselves and the chain is rather abundant in our shop. Gears on the other hand by far are not easy to come by for us and are far more expensive for us to get.
It looks pretty good. I really like how you’ve made everything so simple, with a 2-piece shell. However, there are a few superficial features that I don’t know that you need:
-The height of the module is excessive. More than 5-6 inches is too much. I see you have already elected to fix this.
-Where is the bevel gear and sprocket coming from? If it’s the vex one, how are you making a 3/8" hex bore sprocket?
-Those gear teeth up top look cool, but you may not be able to make them even with a 4th axis CNC. Gear tooth cutters often have a very large radius, so it could be difficult to cut small gear teeth there.
-You need a second bearing up top. Don’t believe me? Stick a shaft through a single hex bearing and see how much it wobbles.
-Consider the new Vex colson wheels; go for something on the order of 2-3" diameter wheels, as the structure of a swerve drive can sometimes allow it to ride over bumps using the abnormally protruding wheels.
-The mounting for the bearing plates near the top look a little unstable, as they are supported on only one side with thin walls. This would be a lot easier if you scrapped the whole tube-shape design.
As a side note, put all your encoders in CAD. It will make life so much easier. Also put motors, mounting, etc. into the CAD.
The height has already been modified in my most recent update to the module (will post photos a little later)
The bevel gear is the vex one and from the model it looks possible to bore it out to 1/2in and then broach a 1/2in hex into it which is what I am hoping for. If not I will need to rethink that portion.
The “gear” pattern at the top is currently meant to interface with an htd 5mm belt allowing the module to turn. I will be making some cuts within the next few days to see if it is possible to make a cog in house that will effectively mesh with such a belt using our resources at hand.
As for the encoders, motors, mounting, etc. those are all things that are yet to come. I have begun sketching some layouts for the other components but for the moment this is what I’ve accomplished these past two days.
I feel like doing maintenance on the worm gears would be a pain, getting them set is probably difficult. Why does there need to be a full sleeve around the whole unit? Weight can be saved by cutting a lot of that out. Also, it looks like you have a belt, but sprockets instead of pulleys, or vise versa. Pulleys can save a lot of weight, which could be extremely beneficial.
If that’s a belt pulley you should be fine with a ball end end mill actually.
If you bore out the bevel gears do some heavy load testing to be sure they won’t fail. If the shaft fails before they do, that’s a good thing.
Encoders always kill me actually. In particular, you may have to put the absolute encoder for rotating the module on the gearbox you use to turn it, rather than on the module itself, and compensate in code.
As awesome as that looks, I cringe when I see all those hardened steel ball bearings running on aluminum. Did you test these modules? Do you not see any pitting in the thrust bearing raceways because of the contact pressure?
A few revisions since my initial post. I’ve added a second bearing to the vertical shaft and the most noticeable change is that I’ve made the shell shorter by 2in. The shell is now only 6in tall. I tried making it even shorter but the change in weight was negligible and the grooves for the UHMW were starting to get too close together.
The “belt” in the pictures are just a representation for the chain. The main reason I went with the chain was it was easier to pack it into a smaller package. It is also easier for my team to work with sprocket/chain than with belts/pulleys.
