Re: Winning Drive Train
 

I love how a question asking what the most effective mechanisms were turns into "talk about what my team did" time.

Simple wins. Always. This is the year that convinced me of that (finally realized which parts of design are OK to think inside the proven box...). You probably won't see any more crabs out of me.

Looking at images of the total winning teams, I'm seeing a trend in simple drive train. This is the first year in a few that we've had a crab make it to the finals on Einstein, and come out on top. Two thirds of the winning alliance was 6 wheel drive.

Simple rules the day. Always.

Re: Winning Drive Train
 

OK.

We drove wide, skid steer with a purposely shortened (14" center to center) wheelbase so that we would have less scrubbing while turning. With our CG just forward of the rear set of wheels this gave us excellent turning and control without reducing the standard ~15 LBf thrust allowed by the rover wheel CoF. Interestingly enough, we're not tippy either ... although I wasn't to worried about tipping with the low CoF this year.

We used a single CIM on each side, directly (through a toughbox) driving the rear wheels and chained to the front wheels. This gave us a safety margin should we throw a chain (not expected ... but one of the highest DFMEA values)

Additionally we gave our driver the ability to control the power curve to the wheels and thrust limiting, this allowed him to control wheel slippage as needed.

Next we added two 16" triblade propellers which added ~4.5 LBf each. These were reversable (although less efficient in reverse) and assisted us in direct line (forward and reverse) as well as turning. With this extra thrust we would overcome the trailers friction, thus the tail (trailer) never wagged the dog (robot).

While thrust for both the propellers and drivetrain were driven off the same joystick, they had different power curves. This allowed us to tune each independently.

The above gave us what we were looking for, in our strategy, The ability to get off the line quickly and the ability to pin (in both autonomous and teleop).

Re: Winning Drive Train
 

That's a neat feature of this year's game...with limited traction, we decided early on to not spend much time on the drivetrain, but put our effort into making the robot pick up balls quickly and unload them quickly.

Some qualitative data....we used the kit drivetrain, completely, one CIM per side, toughboxes, two wheels per side, kit sprockets, bearings, axles, spacers, and chain. But our driver noted after watching videos of our robot in action: "Wow, we're slow!" so we ordered some 20 tooth sprockets to replace the kit 15 tooth transmission sprockets, and made the robot faster in Atlanta.

Re: Winning Drive Train
 

In the course of testing, we found that the difference between the static and dynamic coefficients of friction between any given rover wheel and the regolith was about 20%. However, the difference between clean and dirty (covered with FRP dust) wheels was on the order of 40-50%. I believe that this accounts for some of the discrepancies between the published CoFs and values that multiple teams found empirically in the first week of build.

So while we had some sort of software traction control implemented at all three of our events this season, it was our "mechanical" traction control that helped us the most. Each of our four drive wheels had a paintbrush mounted against it at the top of its rotation. These brushes did a superb job of keeping our wheels clean during and in between matches, and spared us the chore of having to diligently clean our wheels by hand (though we did need to periodically clean out the brushes - you wouldn't believe how much dust they would accumulate). As a result, we were typically a very slippery target for opposing robots (look at our DPR stats this season, although in Philly a couple of problems made us a sitting duck once or twice). Multiple other teams commented on our ability to break out of pins, and it was only when it was 2-on-1 that we were really immobilized.

However, it is interesting to note that our choice of wheelbase meant that despite our tractive advantages, we weren't a very good pinning bot. We were a "skinny" bot because of dimensions of our scoring system, and we had a 4WD setup on a shortened wheelbase towards the rear of our machine.

Top view:
As a result, if we were pinning another bot, it was easy to turn us because of the distance between our robot front and our wheels (our chasis acted like a long lever). Thus our pins were usually not hard to get out of.

Lastly, we found that TOO much traction control was not always a good thing. Our initial software (used in San Diego and Philadelphia) fought to keep the bot on course and slip-free, making drifting and the effective use of the trailer's momentum to quickly turn more difficult. In Atlanta, we toned it back just to acceleration limiting, and our driver responded with our best robot performance (despite the outcome) of the year.

Re: Winning Drive Train
 

More wheel's didn't give move friction. You spread out the weight across the wheels. Traction control was a way to get the theoretical maximum acceleration. we did a crab/swerve drive.

Re: Winning Drive Train
 

I will counter this and say that this year, more wheels did actually return more traction. I know that physics says that Friction force = coefficent*normal force. However, this is for a perfectly ideal surface. This year, I witnessed teams with 8 wheels either in a dually 4wd configuration or in a standard 8wd configuration, have more traction than a similar 4wd or 6wd drivetrain. Our alliance partner at Peachtree utilized a dually 4wd layout with 8 total wheels, and they accelerated faster than us, with 6wd plus fans and follower wheel based traction control.

If I were to redesign a drivetrain for this year, I would do either a 4wd or 6wd layout but with dually wheels, and have dual ducted fans as well.

Re: Winning Drive Train
 

you witnessed it? or you measured the actual amount of force each drive was making, and compared them?

Re: Winning Drive Train
 

Although, I do not disagree with your general premise of simple wins, there has been a swerve in the finals of Einstein each of the past three years and two of them have been on the winning alliance. Having said that, that means that 3/36 teams on Einstein the past three years have had swerve drive and 2/9 of the Champs winners, simpler drive systems still far outnumber the more complex ones.

148 had a swerve last year although it wasn't the same as what teams typically think of as crab. 71 made it to the Einstein finals in 2007, although they did not win.

Re: Winning Drive Train
 

These are all cases where it's really hard to separate cause and effect.

In 07, assuming the rest of the robot was the same, 71 probably would've had equal odds of making einstein if they had gone with simpler drive.

In 08, even though the crab was much more a part of their strategy than in 71's case, 148 probably had similar chances of making Einstein with a small, fast 4/6wd with ackerman steering (or just straight tank).

Now, in terms of comparing acceleration, traction, etc... You CAN'T do it by watching video or matches. period. There are too many variables that you can't account for. Different drivers, different software, etc...

If you want to make any case for what driver pushed harder or steered better, you have to have some sort of data to back it up. The 4wd team XXX had pulled with 12 lbs of force and our 72 wheel drive pulled with 12.4 pounds. Our 4wd turned at 176 degrees/second with a trailer and their 72 wheel turned with 165 degrees/second.

Now, I'm pretty much 100% certain no one has done this across multiple teams, and we can't trust numbers from different teams to compare (differences in measuring equipment and floor would skew it too much), so we can't really go around making these claims can we.

Re: Winning Drive Train
 

I believe the teams won more due to the fact that their drivers and strategies used the crab/swerve to the best of their abilities. Rather than using crab to drive around and strafe like typical bots did, they found a way to use the uniqueness of their drive train to perform a unique task.

148 in 2008 played ridiculous defense and could get through tough spots that others couldn't. That was key in most of their games on Galileo and Einstein. 111 in 2009 immediately strafed to pin an opponent's bot AND position their own bot for top loading. Al mentioned the dynamic centers of rotation (in the crab/swerve thread) that all crab drive trains discovered this year, which was key to their success in this maneuver -- they wouldn't have been as effective in that autonomous strategy with skid steer.

Re: Winning Drive Train
 

I disagree. The reason 111 was such a good dumper was because of swerve; it simply was not possible to lock them down 1 on 1. They were also able to slide sideways and keep the front of their robot oriented toward the trailer. If there was any one thing I could have changed about our robot this year, it would have been the addition of swerve drive (besides, we actually came at weight this year...)

If you watch 1717's videos, they had a trick where they could go face to face with another robot, then slide around to hit them on the side. This is especially useful because for robots like ours and 1717, if we get you on the side we will fill your trailer. Unless you have fans or swerve, it simply is not possible to break out of one of us pushing the side before we empty our screw into the trailer.

Re: Winning Drive Train
 

we just used 4wd tank steer.....

we seemed to be able to push people around with no problems....

though, I definately felt teams with other drivetrains(fans,swerve..etc..) had more of an advantage........

Re: Winning Drive Train
 

I've been following this thread and have notice that, as Adam and JVN pointed out, a lot of this is speculative and isn't being backed up by actual numbers. Allow me to break tradition and share some of our testing methods and summarized results...

When we first started brainstorming we we very focused around being as simple and easy to repair as possible. We wanted no headaches at competition. Therefore we couldn't find a good way of justifying investing the time in developing and producing a crab/swerve drive. INHO those kind of systems need to be developed over the off-season not in the time precious 6 week build season. So the first question was: 4 or 6 axle drive...

4 axle vs. 6 axle Testing
We took our 2008 robot, made rover wheel hubs to fit its 7/16" hex shafts. This became our test platform since we could test 6 axles or remove the center wheels and test 4 axles.

We set up a 8' x 24' FRP track. We did a few different tests:
A) Straight Tests
B) Turning Tests
C) Push/Pinning Testing

We ran the robot down the 24' end of the FRP 10 times, each time on a different area so the FRP wouldn't wear in and give us more traction. We timed each run at full speed so theoretically the wheel slip time would be the same. We found that the straight-away times between 4 and 6 wheels were pretty much the same.

For our turning tests we would go from a straight away, into a turn and time how long it took to make a 180 degree turn. Then we would time how long it took for the driver to regain control in a straight-away after making the turn. We found that the 6 axle version made the turns quicker, but took longer to regain control after making the turn. The 4 axle version made much wider turns, but took no time to regain control after making the turn.

For our Push/Pinning Test we moved the FRP over a large piece of carpet. We set up a wall made of tables. We put our 2008 robot (with rover wheels) in different configurations on the wall (front two wheels on carpet, one side of wheels on the carpet, etc). This wasn't so much of an objective test as much as it was a scenario test... We found that the rocking motion of the 6 axle robot hurt it in pinning situations, especially when the robot was pinned front or back against the wall. The rocking motion meant that there were times when you would potentially have no wheels on the carpet and therefore lose any chance you had at getting out of a pin.

This led us to go with a 4 axle robot. Then came the question of how many wheels to use....

Breaking Force Testing for Multi Wheel Axles
We actually did push/pull/breaking force tests for the rover wheels on FRP. We created a jig that allows us to test 1, 2, 3, 4, 5 and even 6 wheels. We locked the wheels in place so we could test the breaking force the jig. We weighed the jig down with ~40lbs (the amount of weight per axle on a 4WD machine). We dragged the fixture with a fish scale across a section of FRP. The weight measured at the point where the wheels started dragging would be the breaking force.

We found that each wheel added to the fixture greatly reduced our breaking force. 2 wheels cut our breaking force in about half. After seeing it continue to drop after testing 3 wheels and 4 wheels we decided to stop adding more wheels since it was obvious that more wheels per axle meant reducing our breaking force.

After further research on running multiple wheels per axle it turns out that Trucks use them to reduce the load per wheel, not for traction purposes. Think about it... Semi-trucks are the most common vehicle for running 'dualies'. This is because each wheel on the axle reduces the load on the wheel. And if you notice, they're typically only running dualies on the trailer end (where a vast majority of the weight it located) of the truck because they want the traction on the cab.

This is what led us to going with a 4 axle, 4 wheel robot.

Re: Winning Drive Train
 

Thanks!

We have a not very old tradition of sharing most of this type of data during build season....and we haven't found any down side to it yet.... :)

Re: Winning Drive Train
 

3who?