What does it take 6WD/8WD to be successful?

[Ken Streeter]

Originally Posted by dtengineering

We've also found that using seperate control sticks for Forward/Reverse and Left/Right can help tracking. That way, when using "arcade" style driving (rather than tank style, where you control each side or the robot with a seperate stick) when you want to go straight forward you just lift your thumb off the "turning" stick, and you will always get a perfectly straight forward/reverse input to the robot. When you want to stay still and turn, you take your thumb off the forward/reverse, and all you get is turn.

Jason,

I'm looking for a little more detail on what you mean by this. I presume when you refer to "taking your thumb off" you're using a gamepad for driver control, rather than the traditional kit joysticks.

Most teams use the two kit-of-part joysticks to control the robot in "tank drive" style where only front/back on the sticks controls anything. In this way, the left stick controls front / back on the left side of the robot, and the right stick controls front / back on the right side.

The other common alternative is to have "arcade drive" where one stick controls the robot via front/back and right/left stick movements.

What I understand you to be suggesting is to instead use the two joysticks so that one of the joysticks controls front/back of the whole robot (like front/back in "arcade style") and the other joystick controls right/left turning (like right/left in "arcade style") but using 2 joysticks to provide this input rather than just one input. Am I correctly understanding what you are suggesting? Presumably, also rather than actually using two separate joysticks, you simply use a gamepad to provide the above control?

Our team has explored a lot of different drive approaches, but I don't think we've ever tried the above. Sounds like something our team should give a go during the off-season... Thanks for sharing!

--ken

[Chris is me]

Originally Posted by Ken Streeter

What I understand you to be suggesting is to instead use the two joysticks so that one of the joysticks controls front/back of the whole robot (like front/back in "arcade style") and the other joystick controls right/left turning (like right/left in "arcade style") but using 2 joysticks to provide this input rather than just one input. Am I correctly understanding what you are suggesting? Presumably, also rather than actually using two separate joysticks, you simply use a gamepad to provide the above control?

I thought this was actually pretty common, driving with an arcade mode, but with separated joysticks. It's something I've certainly wanted to try for awhile, but I'm stuck on teams that love their joysticks and tank drives.

Maybe if my team enters a mentor match I'll bring my gamepad...

[dtengineering]

Good point, Ken... yes, I was talking about using a PS2/Xbox 360/etc. style game pad. We've been using them for three years now and find that they offer more flexibility for the driver than the joysticks did.

The primary benefit is that the driver is able to move around freely while driving... an advantage when something is blocking his or her view, or... as in Overdrive, when reflections make it difficult to see parts of the floor. It is also useful for testing in the shop as it is easy to pass control from one driver to the next and to position the control board on surfaces of differing heights, or even on the floor, while allowing drivers of different heights to drive comfortably. It also makes it easier to design a compact control board, as the game pads don't need to be fixed in place and can easily stow inside the control board (we build ours as a wooden box with a hinged lid) as opposed to being exposed on the outside of the board. Minor benefits, perhaps, and ones that could be duplicated with joysticks with some effort, but benefits nonetheless.

We've experimented with tank drive, and while it definitely offers some advantages when it comes to turning about one side of the robot or the other, our drivers have preferred "arcade" style with the turning and forward/reverse functions seperated. I wouldn't say that it could be proven to be, objectively, the very best way for everyone to drive their robot, but it has worked well for us.

I guess the main point that I was making was that once a team gets a drive train that is mechanically capable of moving and turning that there are still optimizations to be had through sensors, software and driver interface in order to maximize the "successfulness" of the drivetrain, even for the relatively simple 6wd/8wd layout.

Jason

[buildmaster5000]

Originally Posted by dtengineering

We've experimented with tank drive, and while it definitely offers some advantages when it comes to turning about one side of the robot or the other, our drivers have preferred "arcade" style with the turning and forward/reverse functions seperated. I wouldn't say that it could be proven to be, objectively, the very best way for everyone to drive their robot, but it has worked well for us.

I guess the main point that I was making was that once a team gets a drive train that is mechanically capable of moving and turning that there are still optimizations to be had through sensors, software and driver interface in order to maximize the "successfulness" of the drivetrain, even for the relatively simple 6wd/8wd layout.

Jason

I am just going to add that while hardware is important, we had an instance at the DC regional this past year where our driver was complaining that he couldnt drive straight, but is able to overcome it. After the season, I read somewhere that the library for Jaguars has an output from [-.9, 1.0] which caused the robot ot turn a little as it was directed to drive "straight." We were using joysticks, no gyro, and the default WPILib for anyone who wanted to know.

[JesseK]

I'd like to add that the frame cannot be warped**, regardless of "6WD", "8WD", "Rocking 8WD", "4 traction + 2 omni", etc. This is one of many key elements of any successful drive train. Our team defines "success" as a drive train that moves at a speed we've set forth in our strategies and with the [strategically] desired pushing power while never breaking down.

Oh yea, reliability. The drive train can be creative or simple, yet if the drive train doesn't move then the design of the entire robot is for naught. Not only do teams need testing to find/fix failure points, they need a maintenance plan that quickly exposes any failures that occurred during previous match play. If you have a choice to fix a kicker or your drive train 30 minutes before a match -- choose the drive train.

**Teams that do custom welded frames need to ensure the frame is flat and 4 wheels touch the ground when 4 wheels are supposed to touch the ground. A good aluminum welder or welding mentor is a MUST for this reason. If only 3 wheels are touching the ground when there are supposed to be 4, then the bot will skew one way or the other when put on carpet. Not only does this make a perfect autonomous nearly impossible, it makes driving the robot a bit of a hassle in some situations.

[Andrew Schreiber]

Originally Posted by JesseK

I'd like to add that the frame cannot be warped, regardless of "6WD", "8WD", "Rocking 8WD", "4 traction + 2 omni", etc. This is one of many key elements of any successful drive train.

Teams that do custom welded frames need to ensure the frame is flat and 4 wheels touch the ground when 4 wheels are supposed to touch the ground. A good aluminum welder or welding mentor is a MUST. If only 3 wheels are touching the ground when there are supposed to be 4, then the bot will skew one way or the other when put on carpet.

Excellent point, I point you to one of IKE's threads for more information on frame flex and its effect. http://www.chiefdelphi.com/forums/sh...ad.php?t=68307

Edit: And because it hasn't been posted yet. WARNING MATH AHEAD! (This is where I need a smiley that is a big warning light) http://www.chiefdelphi.com/media/papers/1443

[IKE]

Somewhat contrary to my post, flex isn't always a bad thing though. It can make it tougher to turn, but it also can make it tougher to BE turned. You do want rigid rails along your chain paths though to make sure chains do not jump.

The nature of the kit 6x6 will give you a chassis with pretty good rail stiffness, and a pretty soft torsion to it. This will want to cause opposite corners to dig in when you are turning (bad), but also causes the same corners to dig in when someone is trying to push you around (this can be good if you are trying to hold your ground). You could probably do a pretty good master's thesis on torsionally rigidity and its effects of dynamic behaviour. I know I have read a dozen or so SAE papers on the subject.
**************************
Again though the most important aspect is to understand your functional objectives and priorities. Initially we ranked our mobility objectives as:
Miving is given.
Ability to go over bumps
Ability to hold ground while shooting
Ability to maneuver well
Ability to move long distances quickly

This year we knew we wanted to clear the back row, and move to the midfield, thus the high priority on going over the bump. (In theory you can have 5 balls in a line that the opponent cannot stop you from scoring. That plus hanging equals 7 points which could win most matches.) In 2006, we were pushed around while trying to shoot, thus the hgih priority on holding ground. After the first competition, we found that more defense was played on the goals and balls than robot to robot and thus re-prioritized maneuverability over holding ground. Thus we changed from grippy wheels to the eventual migration to Omnis in the rear (we also needed less rock for ball collector improvements).
Understanding your objectives/priorities/assumptions helps you make decisions when you need to make improvements.

Amount of Rock, which wheels, and how you construct your frame are secondary to how you want it to ideally behave, and where you will make compromises. We consider these design attributes and/or tuning elements used to get the behaviour you want.

(P.S. I really liked the architecture of 201. If we didn't have all the experience and capability with 6x6, I would have pushed hard for an identical architecture. With a little more power, it would have been even more lethal.)

[AustinSchuh]

Originally Posted by Ken Streeter

What I understand you to be suggesting is to instead use the two joysticks so that one of the joysticks controls front/back of the whole robot (like front/back in "arcade style") and the other joystick controls right/left turning (like right/left in "arcade style") but using 2 joysticks to provide this input rather than just one input.

This is how 254 and 968 have driven their robots for the last 2 years.

On 971, we do the same thing, except we use a steering wheel instead of a joystick for turning.

It's nice for the exact same reason that Jason described, you have decoupled the steering from the throttle. I don't know if this is common or not, but we decouple our steering from our throttle in software as well. Our steering wheel controls the radius of our turns, not the turn power. I've found that this makes it a lot more accurate. If you like how your current turn trajectory is going, and just want to go slower, all you have to do is let off the throttle, compared to letting off the throttle and the steering power. Fiddling with this kind of stuff is a great summer project. It's amazing how much more responsive good code can make a robot feel.

As a bonus, read up on IBM's Trackpoint mouse. There's a pretty cool feature that it has that helps with robot handling.

[ajlapp]

what does it take to make a great 6 or 8 wheel tank drive?

Chain/Belt management
Unless you plan to go gear to gear, a la frc25....then you need to plan out how to properly tension your chains. The West Coast chassis and its derivatives all have dedicated chain tensioning designed in.

Force at the Carpet
Great turning requires the proper amount of force at the carpet. Drop center can effectively shorten your wheelbase and help your cause, but without enough force at the carpet you'll get poor turning. Do some quick calculations or use the JVN whitepaper to understand how wheel diameter and subtle gear ratio adjustments can lead to huge leaps in the force available for turning.

Suspension or Flex
Chassis flex or dedicated suspension setups may be the secret to amazing 6wd and 8wd chassis designs. 254 and 968 deploy lots of drop and small diameter large force generating wheels to get the job done...they are also careful CG managers. But it seems to me that the sheet metal bots and those who have special suspension features tucked inside also demonstrate excellent turning prowess. frc118 has deployed suspension in the last two seasons.....I've heard nothing but positive things.

[TEE]

Hey, everyone who contributed, thanks a ton! I look forward to putting this information to use!

Also, thanks (primarily) Jason, for the recommendation regarding programming and control.

[Jared Russell]

Originally Posted by JesseK

I'd like to add that the frame cannot be warped**, regardless of "6WD", "8WD", "Rocking 8WD", "4 traction + 2 omni", etc. This is one of many key elements of any successful drive train. Our team defines "success" as a drive train that moves at a speed we've set forth in our strategies and with the [strategically] desired pushing power while never breaking down.

Actually, one of the many things that an all-traction wheel 6WD or 8WD has going for it is that it tends to be far more robust to flex, uneven terrain, or damage than those with alternating omni wheels, all omni wheels, swerve drive, or all mecanum wheels. As long as at least one wheel on each side is in contact with the ground, you can still drive straight and turn (though how well you turn and about which point you turn depends very much on which wheels are touching). We had rounds where we took significant frame damage or lost one or more chains and were still able to drive and score effectively (with an all-traction, drop center 8WD).

Obviously, things always work better with a pristine, square chassis, but the drop off isn't as bad as it would be with other drive types.

[Daniel_LaFleur]

Originally Posted by TEE

Basically what the title says... what does it take to make a great 6 or 8 wheel tank drive?

Here's my requirements for a good drivetrain (take it for what it's worth):

1> It must be controllable -- Having a robot that can go 20'/sec doesn't help if it's uncontrollable.

2> It must satisfy the games objective and teams strategy -- Being able to push a tank across the field does no good if there is no tank to push. Look at the game and its requirements and design for that.

3> The controls must be intuitive -- The driver will be ineffective if he has to constantly look at his controls and not the playing field.

4> It must be reliable -- It makes no difference if you have the greatest crab/swerve/omni/14-wheel drive if it is constantly breaking. The kitbot drivetrain is reliable enough to take you where you need to go.

5> It must be maintainable -- Things happen, and even the most reliable drivetrain can be damaged. If you cannot fix it during the competition (IE fast) then you need to rethink some of what you are doing. Unmaintainable leads to uninspiring.

You'll note that I did not discuss type of drivetrain, this is because that is very game specific. The best drivetrain is the one that accomplishes the job at hand, regardless of its configuration.