pic: Plywood Butterfly Drive Render

Here’s a render of a butterfly drive I’ve been designing for fun and Solidworks practice.

-Structure is laser-cut .25" birch plywood
-Only other custom components are 3d-printed spacers and axles turned from hex shaft.
-Currently weighs 30.5 lbs.
-Geared for 5.7 fps on Colson wheels and 11.5 fps on omni wheels (adjusted).

Questions or suggestions are welcome.

Is there any reason why you cant shift the colsons closer together and use two pistons between the two sides instead of the four to reduce complexity? Otherwise it looks good although you may find you want to gear it to be faster.

I would highly suggest you have the traction wheels on the outside and the omni wheels on the inside. We did it like your current setup this year, against the advice of teams like 148, and we regret it. When you shift into traction and want to push you’ll tip slightly and end up partially on your omni wheels.

So butterfly drive is like octocanum with omni wheels instead of mecanum wheels?

Looks good, I want to see the real life prototype!

I expect this to depend a lot on the wheel base of the robot while in traction mode (distance from center of front traction wheel to center of back traction wheel). It probably depends on the torque the piston provides on the wheel modules as well. What was the wheel base for your design? Did the tipping result in any lifting of the opposing robot? Curious on your results… for science.

My biggest pet peeve of this type of drivetrain approach will always be that the two CIMs on the front don’t contribute much power in a pushing situation. If you are geared to be traction limited, I don’t think this is an issue.

What about arranging the power transmission components like this? (Warning: MS Paint Art):http://www.bitbuckets.org/wp-content/uploads/2014/06/Butterfly-Drawing-e1403055352939-1024x405.jpg
That would allow all 2 (or 3) CIMs on each side to drive both wheels and move the omni wheels to the inside.

I like it. I have to ask, does your team have a laser capable of making this? If so, what kind?

We work out of a local hackerspace with a Trotec Speedy 300. Some of the parts are longer than the laser cutter’s bed but can be cut in two operations by turning the wood sheet around.


This is pretty close to what we will be prototyping in the fall. Though we will have spots for at least 4 CIM motor mounts per side.

I have been thinking of a way to have the CIMs still face towards the outside of the rail yet recess into that space between the modules and distribute power like your drawing. This year we used the space on one side to hold the cRIO while on the other it was just wasted space. The reason I would want the CIMs to face the outside is it would allow the overall rail thickness to be reduced since it wouldn’t have to contain the entire CIM body.

Here’s a 6-CIM version, with the wheels belt-driven from a central gearbox. I did not switch the omni and Colson wheels; the robot has to tilt to an extreme angle for the raised omni wheels to touch the ground. It now has adjusted speeds of 13.8 and 6.9 fps.http://www.bitbuckets.org/wp-content/uploads/2014/06/butterfly6cim-1024x486.jpg

It seems like you can literally flip your modules around to put the omnis on the inside and the traction on the outside. Just move the shaft with the pulley on it a bit closer in, flip the modules, and you’re gold.

Looks good! Always love to see laser cut wood construction in FRC. Really like the 6 CIM setup with the reversed gearbox. Lot more open in the middle than a lot of octocanum and butterfly designs. With the 6 CIM layout especially, I would recommend flipping your wheels so that it pivots about the traction wheel rather than the omni. Doing this prevents the module from being side loaded when pushed sideways in traction mode.

Have you considered using pancake cylinders for module actuation to save some space and weight?

Looks great! In addition to the above (swapping wheel placement), I think you could get a more compact design using four fat pancake cylinders instead of four longer cylinders with a long lever arm on the butterfly. You’re losing a lot of pod-turning torque based on that angle anyway, so why not mount a pancake cylinder so it can push the pod straight down?

(We’ve been iterating octocanum for quite a few years now, and that’s how we intend to do it this year if we keep octocanum this year…which we may not.)

The biggest issue with this is it makes it harder to do your reductions. If the omni wheel is floating, it’s the module that normally is powered first before the traction wheel. In many designs this means the CIMs need to float with the modules, 1477 did it this way this year. Floating the traction wheels allow for much easier gearing. We didn’t see any bending in our modules this year that had the traction wheels floating but we used steel side plates. I’ve heard of teams putting delrin blocks between the modules and the frame to prevent the modules from taking all the side loads when the traction wheels are pushed down.

It’s not too hard with the central gearbox design since you can do a second reduction with the belts to the modules. From JVN’s calculator, a 12:60 reduction in the gearbox and a 24:42 reduction with the belts gives adjusted speeds of 6.4fps on the Colson wheels and 12.9 fps on the omni wheels.

I’ll look into doing that.

I noticed a few problems with my design that I also need to fix in the next iteration: the belts to the modules run into the bellypan, the bolt that serves as an axle for the gearbox is not adequately supported and the bottom two CIMs in each gearbox can’t be installed or removed once the chassis is assembled.

The other reason we prefer to float the traction wheels is that if we lose air, or if we lose a belt in a module we default to the omni wheels which is where we spend around 90% of our time anyway.

If it simplifies the design, you could consider the option of using shifters for the central gearboxes rather than a pulley reduction between the traction and omni wheel. That way, you can shift from high speed to high torque independently of which wheels are on the ground and pivoting about the traction wheel isn’t a big hassle. However, this is added complexity.

It’s great to see another team going with this method of construction. Mine has been doing laser cut plywood chassis in the past two years, and we love it.

Is there anywhere I could find more information on your designs and construction method? What thickness and type of wood did you use? How did it hold up in competition?

It’s hard to find high-quality information about wood chassis in FRC since so few teams have used them.