We had kicked around a swerve last year until we went to CMP and saw how much machining it took to make a proper one. This year necessity drove our design like it always seems to do 
We decided that tank would be inadequate this year, so we CADed up parts and CNCed them on our 20+ year old machine.
The amazing thing is, that we really only started assembling at the end of week 3 I think
We decided in our initial brainstorm that swerve would be the way to go. After that we CADed our brains out and sent out parts to be water jet cut and welded for our frame in week one (Thank you Omax). The modules 4"x4" (1/4 W) was cut into segments and sent to another welder along with the 1/4" al. plate to be welded into the modules initial shape. From there they were taken to a new machining sponsor (Puget Sound Precision) and were machined on their 9 axis CNC mills. We got them back and assembled them during week three and were driving at the end of week 4
Nice. We had the plates CNCed in our small little shop out of 1/8" plate. It took a long time.
waitâŚtank doesnt work? did i miss a memoâŚ
Tank works, we just decided that this year would be a good year to start swerve, and that it would help a lot with the trailer.
Big1BoomâŚtheir modified swerve is what we call a twerveâŚnice job guys, thatâs a very creative idea. i had a similar idea last year but have never been able to actually implement it. If i may ask, how is that working out for you perfomance wise?
Surely you are running closed-loop position control for your steering motors. What type of sensor are you using on them?
we had the same problem with our first swerve drives, one somewhat simple thing you can do is to move the chain and sprockets connecting the wheel axle to the upper axle to the outside of the module. As long as this doesnât cause your module to become wider then the diameter of the wheel it shouldnât hit anything. this makes tensioning and replacement of this chain much easier, also it opens up more room on the inside of the module.
http://www.chiefdelphi.com/forums/showthread.php?t=66234&highlight=swerve+module
This is our teams last swerve design, one other thing we did as far as maintenance was use hex axles and snap rings to make everything very easy to change. Our '08 bot,which used these modules,never broke in competition but it was way better than '07 when we had to get 4 people and sit for an hour to get the wheels out and replace the tread, now it only takes 1 person 30 minutes or less to replace all 4 wheels.
I think all of the wheel modules Iâve seen online have had their main housing either welded together or machined from solid.
Our modules are bolted together using aircraft-grade 6-32 machine screws.
http://www.chiefdelphi.com/media/photos/32573
Donât get me wrong though, Iâm a big fan of welding. It just seems to me that it would be easier to do service work on a given wheel module if it could be taken apart. Using screws to fasten the parts together does offer a serviceability advantage IMO.
I also believe that the weight of bolted housing is comparable to the weight of a welded housing - our modules weigh less than 4lbs each.
Still, the best wheel module is one thatâs so reliable it doesnât need to be serviced very often, or even at all!
Performance is quite satisfying, we are able to use crab drive and âwarthogâ swerve and transition between the two seamlessly. BUT, we can and do also use simple tank differential drive for rotating in place as it seems to be the best for that.
We are using seperate PID control for both front and back steering. As for the sensor, we are using an unlimited turn potentiometer. While this does have a 20-degree dead band, we oriented it in such a way that the dead band is facing the rear of the robot. When we near the dead band, the swerve module flips orientation and reverses power so that we can move in that direction. The âflopâ is extremely fast and is nearly seamless. While we would prefer to have the truly seamless feedback that an absolute optical encoder would provide, the two eight dollar a pop pots were MUCH more affordable than the 350 dollar optical encoders were. Next year we plan to utilize two pots per wheel and align them in opposition to effectively eliminate the dead band.
After this year I think running the chain on the outside of the module will taken into consideration with high necessity.
We have had ZERO problems with the modules themselves, the steering chain runs have been the problem children. However, thankfully we were able to identify and solve this problem after 3 days on our practice field prior to ship. PORTLAND HERE WE COME!
$350 eh?](http://www.usdigital.com/products/encoders/absolute/rotary/shaft/ma3/)
We used those on our KingKrab, and so far have had no issues. 111 and some teams have reported some possible issues with static discharge however.
We used rotary encoders (non-absolute) to close a position loop around our steering motors. Our modules can twist in any one direction indefinitely.
To align our wheels upon robot enable (to tell software where âforwardâ is so we can measure displacement from zero) we use a pair of IR emitters/optical sensors. The wheels automatically snap forward (it takes .5 seconds) and then normal operation begins.
Itâs not the best solution⌠two separate sensors + wiring and a pre-operation software routine, but it works.
theyâve been used on every 1625 swerve, we started also replacing pots elsewhere with them. Never had the static issue 111 had with them though
so another vote for ma3âs
I stand corrected thanks for the tip, I assure you that it will not fall upon deaf earsâŚ
You are right, Dave! We will use this in the future. Thanks Adam.