[trilogy2826]

Small sidetrack first:
2826 has built 11 drivetrains in the four years we have been competing:

Competition:
2009: 3 wheel crab, Wide
2010: 4 wheel longitudinal beam suspension (2 traction in front, 2 omnis in back), Long
2011: Octocanum, Long
2012: 8 wheel, 2 center drop, 5" Colsons, Long

Offseason:
4 wheel independent Crab, 15" mecanum (yeah, 15"), 6" mecanum, 6WD drop center. 2WD with rear casters, Butterfly drive, plus some various prototypes that are more or less bits pieces.

Two things that really stand out from what we've learned:
1. Never use mecanums unless they are for whatever reason mandated by FIRST (sorry to all you mecanum lovers out there)
2. Colson's in a drop center drive make a fantastic drivetrain.

Back on topic...
This last year we switched to Colson 5" for IRI from a custom wheel that we designed. The drive was an 8WD with the two middle axles on the same plane dropped 0.125 wrt the outboard axles. The wheelbase between all axles was an equal 9.25 in. We found that when we moved to the Colsons, there was an excessive scrub when turning resulting (what we believe) from 3 wheels on one side touching momentarily. This caused, to some degree, the same effect that a long 4WD with tractions is known for (hopping, excessive battery drain). We cheated and shaved down the diameter of Colson's on the outer axles to increase our "drop". We ended up with 5/16" vertical difference between the outer and inner axles, even though they were only 9.25 inches from each other.

Needless to say, this bothered the heck out of me, so we dug into it over the last month. I know we haven't yet discovered all the root causes, but these have led us to rethink a few things.

First off, my definition of how to measure "drop": Draw a ground reference line. Draw a line parallel to this ground line between the center of both outboard wheels. The center of the inboard wheel(s) is below this line by your drop distance. This might seem obvious, but I want to make sure I am stating everything following from a baseline.


Root causes of the design failure:
1. Lack of understanding of the geometry: We simply went with an 1/8" drop because our CD research pointed to that being an optimal place to start. In a 6WD, 1/8" makes sense. If you take a symmetrical 6WD with the center dropped 1/8" and you rock the chassis so that the "rear" and middle axle wheels are resting on the ground, the lifted wheel is raised from the ground 1/4"

In the 8WD with the outboard wheels raised 1/8, you first need to decompose this into a mini 6WD. This is easiest to imagine by simply deleting one of the outboard axles

In this case, the effective raise of the outboard axle is only 1/8". This lower clearance, combined with the following, caused our issues

2. Carpet to Colson compliance: On a consistently hard surface like linoleum or concrete, this small clearance works perfectly well, with no hopping. Though FRC carpet is pretty compact, it can flex and compress a significant amount under stress. This added to a full robot weight on top of one wheel (the dropped wheel in this case, on one side of the bot), will allow your raised, non-compressed wheels to touch the ground simultaneously.

3. Dynamic turning compliance of colsons: Didn't know what else to call this. Based on high speed video, we determined that while turning, the Colsons tended to compress even further past their static state mentioned above. We attribute this to two possibilities: increased force on the wheel based on simple dynamics and the thought that the urethane "twists" causing the material to locally compress. While we were not able to definitively measure this effect, based on video while on concrete, we believe this can account for at least an additional 1/32" of compression.

Solution:
We are going toward specifying relative angles dependent on wheelbases instead of just drop. This is also based on just looking at a simpler 3 wheel per side model, no matter the amount of wheels. I plan on putting out a whitepaper soon on this topic, so please shoot holes in my process now.