Two joystick tank drive vs one stick arcade drive

[mickztlin]

I understand that for driving tank drive you can set one joystick to both sides of the drive train, or have two joysticks set to each side of the drive train. In all previous years we have chosen one joystick for drive train but it seems like two would give back more accurate control. So my question is, what is everyones elses opinion on this?

[wilful]

It is really best for each driver to figure out which setup they like best.

Personally, I like two joysticks on a game controller, but in 2011 the driver preferred arcade on a game controller while the driver for 2012 and 2013 preferred a setup using two full joysticks.

[FilthyArgonian]

Two-stick drive seems more intuitive at first, but with a lot of practice, one-stick arcade drive will give the drive greater control of the robot. This only really applies to the standard drive trains as mecanum, octocanum, kiwi, swerve, or any other type of drive will have different options for control style.

[AGPapa]

Originally Posted by FilthyArgonian

one-stick arcade drive will give the drive greater control of the robot.

Why?

[apalrd]

We have used a mixed style algorithm for the past few years. We currently use the 'CulverDrive' algorithm, derived from the 'Cheesy Drive' algorithm published by the Cheesy Poofs. In both cases, there is a 'throttle' and 'wheel' (steering) input which is mixed in software to a left and right command. The Arcade algorithm does something similar, in a very mathematically simplified way.

We then split the mixed algorithm between the two sticks, so the X axis is on one stick and the Y axis is on the other stick. This separates the driving from the steering, allowing the driver to more finely control both (specifically, it prevents quick changes in throttle control from bumping the steering control, and vice-versa). The driver can maintain a steering command and modify the throttle command easily as they are detached.

You can do the same with the Arcade algorithm by mapping the X input to the X axis of one stick, and the Y input to the Y axis of the other stick.

We use Logitech F310 gamepads, as we were not impressed by the quality of the xbox gamepads (specifically, they do not return to the same center very well compared to the Logitech gamepads). Our drivers also prefer the circular stick boundaries of the F310 vs the octagonal boundary of the xbox, and we coded in assumptions of a circular boundary for the CulverDrive.

We found, after much testing, that the two-stick tank drive resulted in 'trinary' (3-position, almost binary) motions. The driver would usually only put each stick in 3 positions: +1, 0, -1. Both sticks would be full forward, and to turn, the inner wheel stick would go to either 0 (driver removes thumb from stick) or -1 (driver does a full reverse spin turn). We tried to fix this on our 2011 robot using an 'arc control' code that would adjust the inner wheel transfer function (so a 0 on the inner wheel would become a +0.4 or so), but this was an intermediate measure as the driver was trained on two-stick tank and we could not switch to a mixed control setup until the next season.

With the separated throttle/wheel, the driver still drives with a trinary throttle, but can now finely control the wheel input independently to gracefully turn. The algorithm also tries to maintain a constant radius turn for a given wheel input independent of throttle (we can adjust this as one of the calibratable 'feel' parameters)


IMHO, if the driver is not already trained, you should not let the driver pick. We initially did this and most drivers picked the tank drive because they were familiar with the robot mechanics and it made sense to them (there was no 'magic software'). However, with practice, all of our drivers who have tried it have adopted our 'CulverDrive', and the Halo/Cheesy style drives as well. We believe it's WAY easier for a driver to control a high speed FRC robot well with a mixed control, YMMV.

[JohnFogarty]

Originally Posted by apalrd

We have used a mixed style algorithm for the past few years. We currently use the 'CulverDrive' algorithm, derived from the 'Cheesy Drive' algorithm published by the Cheesy Poofs. In both cases, there is a 'throttle' and 'wheel' (steering) input which is mixed in software to a left and right command. The Arcade algorithm does something similar, in a very mathematically simplified way.

We then split the mixed algorithm between the two sticks, so the X axis is on one stick and the Y axis is on the other stick. This separates the driving from the steering, allowing the driver to more finely control both (specifically, it prevents quick changes in throttle control from bumping the steering control, and vice-versa). The driver can maintain a steering command and modify the throttle command easily as they are detached.

You can do the same with the Arcade algorithm by mapping the X input to the X axis of one stick, and the Y input to the Y axis of the other stick.

We use Logitech F310 gamepads, as we were not impressed by the quality of the xbox gamepads (specifically, they do not return to the same center very well compared to the Logitech gamepads). Our drivers also prefer the circular stick boundaries of the F310 vs the octagonal boundary of the xbox, and we coded in assumptions of a circular boundary for the CulverDrive.

We found, after much testing, that the two-stick tank drive resulted in 'trinary' (3-position, almost binary) motions. The driver would usually only put each stick in 3 positions: +1, 0, -1. Both sticks would be full forward, and to turn, the inner wheel stick would go to either 0 (driver removes thumb from stick) or -1 (driver does a full reverse spin turn). We tried to fix this on our 2011 robot using an 'arc control' code that would adjust the inner wheel transfer function (so a 0 on the inner wheel would become a +0.4 or so), but this was an intermediate measure as the driver was trained on two-stick tank and we could not switch to a mixed control setup until the next season.

With the separated throttle/wheel, the driver still drives with a trinary throttle, but can now finely control the wheel input independently to gracefully turn. The algorithm also tries to maintain a constant radius turn for a given wheel input independent of throttle (we can adjust this as one of the calibratable 'feel' parameters)


IMHO, if the driver is not already trained, you should not let the driver pick. We initially did this and most drivers picked the tank drive because they were familiar with the robot mechanics and it made sense to them (there was no 'magic software'). However, with practice, all of our drivers who have tried it have adopted our 'CulverDrive', and the Halo/Cheesy style drives as well. We believe it's WAY easier for a driver to control a high speed FRC robot well with a mixed control, YMMV.

That is really interesting. I'm surprised I haven't looked this up by now. With my old team (1102) I used tank drive because it was more simple to code and also everyone seemed to prefer it because of it's native logic.

Working with my new team and new mentors it was suggested to me to work on coding something very similar to cheesy-drive. I personally love it. The control is much better than anything I drove as a student before. I'm going to have to look up the cheesy drive to see how different it is from the code I've written.

[FilthyArgonian]

when you use two stick tank drive the value from your joysticks might not be identical when you want them to be (due to human error or joystick inaccuracy), it can cause your robot to turn to one side or do things you might not expect.

[FilthyArgonian]

Originally Posted by FilthyArgonian

when you use two stick tank drive the value from your joysticks might not be identical when you want them to be (due to human error or joystick inaccuracy), it can cause your robot to turn to one side or do things you might not expect.

Of course, this is assuming that by two stick tank drive you mean one joystick controls the left half of your drivetrain and another joystick controls the right half.

[AllenGregoryIV]

Anyone that can figure out how to get some variant of Cheesy Drive/Halo Drive working and have their drivers trained with it should do it. We switched to it last year and we will never go back. You'll see a much more smooth motion from your robot once your driver stops trying to turn in place all the time. Even a poorly tuned version of cheesy drive just looks better and can arc around corners much easier. I know some drivers who can get similar motions out of a tank drive robot but it is far more difficult in my opinion.

Also I highly recommend avoiding single stick arcade, since it will make your driver prone to unintentionally arcing the robot because it's hard to keep the stick perfectly straight.

[FilthyArgonian]

Originally Posted by AllenGregoryIV

Anyone that can figure out how to get some variant of Cheesy Drive/Halo Drive working and have their drivers trained with it should do it.

I completely agree with this. Halo Drive is my favorite that i have used (though i have extremely limited experience with regards to robot driving) but in a decision between one stick arcade and two stick tank, i think one stick arcade is better.

[Ether]

Originally Posted by apalrd

We then split the mixed algorithm between the two sticks, so the X axis is on one stick and the Y axis is on the other stick.

I'm wondering if Linear Interpolated Arcade, with the joysticks assigned as mentioned above, would provide similarly improved driver control.

http://www.chiefdelphi.com/forums/sh....php?p=1240254

http://www.chiefdelphi.com/forums/sh....php?p=1329493