Ok, here is a more normal approach to the swerve than my last one. While the other one would not have been used in a FIRST bot, this one could be. When building this one I needed a few things:
- Low cost, perhaps the most important. This would cost about $600-$700 dollars depending on access to CIM,s and which banebot gear box was used.
- Low weight, that which is shown would weigh about 35 pounds. With a little work it could weigh less than 30. I’m currently happy with the weight, but always enjoy taking more out.
- easy of construction. We have access to a CNC router and a small mill but don’t really have anyone who knows how to use them really well (I have self taught myself over the past couple of years). Coaxial swerve was out because right angles are hard for us. The frame is made of 1/2" steel box beam that is welded together and then an 1/8" Polycarbonate sheet that is bolted and several points to the frame for stiffness. I use steel because we don’t have access to a tig welder.
Thanks for all the input on the last one. Any thoughts?
Are you sure you want to commit 8motors to your drive train? ours the steering was all linked so we could tank steer which only took 5 motors you have that option by replacing all of the banebots with sprockets and have 1 motor drive them. I know las guerillas (469) went the 8 motor route and it was pretty crazy to watch.
amazing weight too especially for a swerve drive
I agree with Aren, although we use two Globes for steering. I would stay away from the Banebot transmissions that appear in the steering motors. They might not be able to handle the rough treatment when your robot gets hit from the side or bangs into something.
FIRST off, 
if you need any help with G-Code programming or machining in general just hit me up…
I’m wondering… does that 64:1 gearbox with the mabauchi motor have enough torque to turn the wheel under load?
A suggestion to save weight, for 1/8" polycarbonate you can use a large pop rivet to save weight as well, not to mention the higher population of fasteners will more evenly distribute the forces on the stiffening plate over a larger portion of its perimeter.
Very nice lightweight design, have to say, amazed that all those gears for four wheels only weigh 35lbs. :ahh:
-q
Any particular reason why you ditched the suspension? I think some teams have built suspensions in previous games… (i know a former mentor does but i don’t have the link) I’m pretty sure its possible to do with a couple long pieces of shaft.
Are you taking orders? 
Looks great, especially for the weight. I see other’s concerns about the BB gearboxes under stress… I’m not too convinced thought that it would be that much of a problem (probably something that can be mechanically avoided.)
One more question. You said ‘that which is shown’ is 35 lbs… Does that mean that a set of 4 would be 70 lbs? and does that 35 lbs include the chassis pieces shown?
Great work.
Jacob
I’m using 8 motors in the drive for a couple of reasons. I like the idea of the extra maneuverability and control that is provided, i know that this is more complicated program wise but it is less complex mechanically and is lighter. Two motors weighs less than a chain tensioner plus chain and sprockets. Mechanically it is much easier cause I don’t have to mount everything separately and all the pieces are the same and there are fewer of them.
The motors would be two Fischer Price and two BaneBots (don’t know if we have all these next year so…). I’m pretty sure that the trans could handle the load because there is a bearing at the bottom that all upward force and most lateral force would go into. I would worry a little about the carrier plate but if the modules are well protected I don’t think it would be a huge problem.
I got rid of the suspension cause of weight and the suspension is really complicated. I really liked the idea but it didn’t seem needed for an actual first robot, plus this is much lighter and more importantly costs about half as much. The wheels cost $2.75 each instead of $100 each. And when I say “that which is shown” I forgot that I was only showing half of the bot. The entire thing, modules, motors, frame, everything weighs only 35 pounds. Of course with electronics this could be 10 pounds heavier but that still leaves much more than half the allowed weight to build on top of.
Orders… jee hadn’t thought of that. 
Thanks for the offer of help but I think I can handle it, however if I do need help. If you would like to CAD files just ask.
Thanks very much for all the input.
Alex,
One thought I had after my intitial post is the problem of keeping the steering motors in alignment. It is an issue using two motors and chain for us. To keep all four tracking is going to take some processing power and sensors.
Just out of curiousity, how do you keep the wires on the drive motors from tangeling around the sverve module?
We limit the travel to only one rotation plus a few degrees, max. This year we closed that down even further. The wiring comes right up out of the top of the module and the upper bearing point is a hollow tube which the wires run through.
I was actually worried about how to keep the tubes in alignment. I was thinking of adding some sheets to the frame that would brace between the tubes vertically. And now that you mention it that is another good reason to use four motors, there is again less precision needed. While I do need precision to attach the motor mounts, I can think of an easy way to do that.
In this case since the module turning comes down through the top I can’t run the wires through the bearing. I would run the wires up the inside of the module and then leave a large loop of wire to allow for turning before attaching directly above the module. I would also limit the turning of the module to keep the wire from tangling.
Alex,
One thought I had after my intitial post is the problem of keeping the steering motors in alignment. It is an issue using two motors and chain for us. To keep all four tracking is going to take some processing power and sensors.
We used an “Magnetic Rotary Absolute Encoder”](http://www.usdigital.com/products/absolute-encoders-magnetic.shtml) on each wheel. They act like potentiometers, since they are connected to an analog input and are read as a value between 0-1023. However, they do not have stops like potentiometers and can be rotated freely. Therefore, they have the resolution of a one-turn pot but can be used in multi-turn applications.
We used this little piece of code to keep track of rotations:
char encoder_turns(int pot, int last, int window) {
if(pot - last > 1023 - window)
return -1;
else if(last - pot > 1023 - window)
return 1;
return 0;
}
Since a value read from the encoder will always refer to the same physical position, we didn’t have to worry about loosing positions after turning off the robot. However, since we had the encoder connected to the main (rotating) shaft using two plastic sprockets, we lost calibration there since the sprocket would sometimes “ride on a chain” or skip a few teeth on fast movement. It wasn’t a major problem, since we recalibrated before every round.
We then used 2 PI loops on each module (cw,ccw) to get it in the position we wanted. I could write a whole essay on control theory right here, but I think Jimmy’s post “sums” it up: http://www.chiefdelphi.com/forums/showthread.php?p=617897&highlight=point#post617897
We limit the travel to only one rotation plus a few degrees, max. This year we closed that down even further. The wiring comes right up out of the top of the module and the upper bearing point is a hollow tube which the wires run through.
We also ran the wires upwards through a central shaft. We made them long enough to allow for about 3 rotation each way. We kept track of the turn count in programming and saved it to the eeprom so that we wouldn’t loose track. We also wrote an unspin function that unspins the wires.
Seems to me that window lift motors - or a similar worm drive system - might be better for the steering motors. They rotate slowly - but maybe not slow enough - and have significant torque, and take the bumps pretty well. Plus, they are available cheap in surplus and reasonable in new.
Some newer types sometimes have integrated quadrature sensors (used for detecting objects during auto-close of the windows) that are sensitive and can be used to maintain steering alignment (along with a ‘home’ switch for recalibration).
Some idle thought, is all.
Don
