Moment of inertia and turning performance

[DampRobot]

Our robot was symmetrical this year. Well, not exactly symmetrical, but so much so that our driver has trouble telling one end of the robot from the other. We are using 6wd off a CIM and a FP a side. All 6 wheels are 6in performance wheels with blue nitrite tread, and the center wheels are dropped about an eighth of an inch.

We turn very easily, at least much more easily than most 6wd robots with all traction wheels. However, if you watched our robot play, you would say that it turned very jerkily. Why? The driver and the programming.

[JesseK]

This is merely an explanation for why a LOW c.g. matters.

Consider a c.g. that's off-center to the right and is 18" off the ground (really high for FRC). When moving forward, that c.g. creates a moving inertia. During a sweeping left-hand turn while moving, that inertia becomes a centrifugal force that attempts to lift the left side of the robot. The higher the c.g., the more leverage the force has to lift the left side. This centrifugal force is easily noticable on robots that do high-speed 0-radius turns in place. It's like an off-balanced top.

This lifting causes a major traction loss, even if it only lifts a few tenths of an inch. Realize that with even a few degrees of lateral lift, a wheel that makes flat contact with the field (such as the KOP wheels) will lose nearly all of its traction. Rounder wheels, like the AM pneumatic wheels, will only lose some of their traction. This traction loss is why high c.g. robots cannot do sweeping turns like the more agile robots. A good picture of this concept is on page 41 of this slide deck: http://www.chiefdelphi.com/media/papers/2597

[Nathan Streeter]

Originally Posted by Ether

Segways turn without any* resistance.

A 6WD with omnis at all four corners is like a Segway with training wheels.


*well, for all practical purposes

I'm assuming that you are implying that you do often want a drive train that turns with negligible resistance, and that they can be controllable?

In the "real world" (i.e. with cars, airplanes, segways, etc.) you're definitely right - you basically always want no mechanical resistance in the drive mechanism. As a result, you have a drive mechanism that is capable of being quite responsive. However, if that mechanism is controlled carelessly, it could result in a vehicle that responds easily to accidental maneuvers. These potential side-effects are mitigated by control mechanisms with resistance or "feedback" (such as the yoke or stick on a plane), control motions that aren't inherently twitchy (two hands on a reasonably sized steering wheel), software damping of otherwise twitchy motions (a Wii uses some of this, I conjecture), and/or by training the manipulator to use the responsive mechanism advantageously (fighter pilot).

These are all better ways of getting good controlability out of an otherwise difficult-to-control system; however, I made several unspoken assumptions when I thought what I wrote. I figured teams would be working with control mechanisms that aren't perfect (the current joysticks at least have a little more resistance than the old ones...), that they'd be using control motions that aren't always easy to make steady (unsupported arms on two different joysticks), that most teams wouldn't be using software to damp out twitchiness (perhaps some teams do), and that teams wouldn't be requiring a level of responsiveness that could only be acquired with a resistance-free mechanism.

From our own team's experience, our drivers had a hard time getting accustomed to our very low-resistance drive trains... but we ended up with robots that could drive fairly well by compensating. We didn't want all the responsiveness we had, so we used software to map the distance from the neutral axes exponentially, so that small motions would be less significant. I consider this raising the tolerances, rather than "damping out twitchiness..." which I presume would be feasible but significantly more difficult, and perhaps not even preferable. We also used software so that when the driver hit a button they could limit the robot to half (or quarter) power for fine motions, such as placing a tube on a peg last year or lining up for a shot in 2010. Then we got the drivers to practice a lot.

So, in the FRC robot world, my experience has been that a hyper-sensitive robot is not preferable and that a good way to limit the sensitivity is to leave a little resistance in the system (i.e. put grippy tires at the corners rather than omnis); however, some teams may think otherwise. Perhaps your team disagrees... I could imagine some teams preferring hyper-sensitive robots that are driven well solely by the skill and practice of their drivers...

I apologize for my non-absolute statement that could easily be misinterpreted.

[Ether]

I think your expanded explanation makes a lot of sense.

[Alan Anderson]

Originally Posted by Ether

Segways turn without any* resistance.

A 6WD with omnis at all four corners is like a Segway with training wheels.

*well, for all practical purposes

Have you ever tried to turn an active Segway by manhandling it instead of using the steering control? It has plenty of resistance. Granted, it's mostly due to software controlling the wheels and not to strictly mechanical tendencies.

I think it would be better to say that a Segway has very little wheel scrub to worry about.

[icyplanetnhc]

Originally Posted by BJC

Our team has done a lot of personal testing and evaluation on the top teams in the league and we have found that the CoG and tortional stiffness is generally what sets their robot's drivetrain apart from most others.

CoG, contrary to popular belief we have found that setting the CoG far to one side on a 6wd is best. A 6wd performs best when it is actually a 4wd with 2 extra wheels. Placing the CoG in the middle generally results in the turning point of the robot changing in the middle of the turn as the robot accelerates/decelerates and the outer wheels alernating touching the ground. This is especially important because a 6wd does not turn about the middle wheels but rather between the middle wheels and whichever pair of outer wheels happens to be touching the ground. So by limiting rocking as much as possible performace greatly increases.

Tortional stiffness is absolutely critical to a good 6wd. In 2008 we discovered that without a very stiff chassis the opposite outer wheels will touch down essentially creating a long wheelbase while turning. This will also seriously limits turning.

Edit: I forgot to add, an 8wd has mathmatically much better turning characteristics than a 6wd provided the wheels are properly spaced.

But don't take my word for it, if you don't believe me run some experiments.
Regards, Bryan

Yes, our 8WD drive from last year worked well for the most part. It had difficulty in an off-season game because the softer carpet allowed our front and rear wheels to make contact with the ground. We tried to alleviate that problem this year by increasing the drop of our center wheel.

With regards to method of using a 6WD as a short-base 4WD with 2 additional wheels, this is something that we've considered last year, but I believe we chose against it because it might've made aiming on one side of the robot, either the scoring grid or minibot tower, difficult (this was before we chose to use "alignment legs" at the bottom of the robot). In addition, if our robot's turning axis is biased to one side, other robots may have a large moment arm to turn our robot.

[Nathan Streeter]

Originally Posted by icyplanetnhc

With regards to method of using a 6WD as a short-base 4WD with 2 additional wheels, this is something that we've considered last year, but I believe we chose against it because it might've made aiming on one side of the robot, either the scoring grid or minibot tower, difficult (this was before we chose to use "alignment legs" at the bottom of the robot). In addition, if our robot's turning axis is biased to one side, other robots may have a large moment arm to turn our robot.

While it would make aiming on one side of the robot consistently harder since the point of rotation would be farther from one end of the robot, it isn't actually very different from having the "standard" drop-center 6WD... The drop-center 6WD is, to a degree, just two short 4WDs that are "randomly" switched between. From my impression, the drop-center 6WD rarely actually turns around its center wheels, but is usually turning around a point between one of the sets of 4WD. I would wager that most drop-center 6WDs are actually easier to spin than most "4WD+2WD" drive trains. Since the weight in a 4+2 is more firmly placed on four of the wheels, it seems like it might be more resistant to turning; however, that is conjecture rather than experimentally-verified fact...

Originally Posted by Gregor01

Wrong thread maybe Andrew?

Lol at your story anyway

I'm so sorry! That was the wrong thread! Bad things happen when you have two or so tabs open for CD....

[BJC]

Originally Posted by icyplanetnhc

Yes, our 8WD drive from last year worked well for the most part. It had difficulty in an off-season game because the softer carpet allowed our front and rear wheels to make contact with the ground. We tried to alleviate that problem this year by increasing the drop of our center wheel.

With regards to method of using a 6WD as a short-base 4WD with 2 additional wheels, this is something that we've considered last year, but I believe we chose against it because it might've made aiming on one side of the robot, either the scoring grid or minibot tower, difficult (this was before we chose to use "alignment legs" at the bottom of the robot). In addition, if our robot's turning axis is biased to one side, other robots may have a large moment arm to turn our robot.

Drop definatly is a huge part of the equation. Too much and you rock back and forth, to little and the wheel scrub makes it difficult to turn. This is again why tortional stiffness is SO important. With a flexable chassis you end up with dropped wheels while turning and raised wheels when your not: the worst of both worlds.

As to not wanting to be turned sideways, it seems to be mostly a non-issue compaired to the performance gains, you don't notice the powerhouse teams that use 6-8wds year after year getting pushed around very much. (and when you think about having the CoG on top of the middle wheels in the 6wd wouldn't having a changing point of rotation actually make it harder to line up to something?)

Interesting thread
Bryan

[Michael Corsetto]

Originally Posted by BJC

Drop definatly is a huge part of the equation. Too much and you rock back and forth, to little and the wheel scrub makes it difficult to turn. This is again why tortional stiffness is SO important. With a flexable chassis you end up with dropped wheels while turning and raised wheels when your not: the worst of both worlds.

Any advice on how to maximize torsional stiffness in a frame? Just beef up the gussets? Add a second frame layer a few inches up? I know this is kind of a vague question, considering all of the different frame construction methods, just wondering how teams typically negate this (if at all).

Some of my own observations:

Welded tube frames are always rock solid in my experience. 1662 uses 1"x1.5", 1/8" wall tubing welded frame. No flex whatsoever. This is consistent with many WCD around California.

1678 uses 8020 extrusion frames, and they flex like mad. Maybe it was just the gussets we used, but I have a feeling that the 8020 beams are less able to resist torsional forces. (If I remembered everything I learned in mechanics of materials 4 years ago I might be able to figure it out... ) Last year we had to create a pyramid strut system to our arm apex in order to remove the torsion experience while turning.

How is the kitbot frame's torsional stiffness? I don't have much experience with it.

Awesome thread, all this talk about frame flex has got me thinking!

-Mike

[IKE]

Originally Posted by Michael Corsetto

How is the kitbot frame's torsional stiffness? I don't have much experience with it.

Awesome thread, all this talk about frame flex has got me thinking!

-Mike

Kit bot=Noodle frame. It does have a fair amount of rock, so it works out well.
1114's Kitbot on Steroids improves this using the robust base-plate.

Super structure design can make a frame stiff.
Welding can make a frame stiffer.
Gussets can make a frame stiffer.
Additional rivets at attachment locations.
Gluing.
In general, Triangulation beats Boxification (I made the last one up).

There are lots of ways that can make a frame stiffer, but many do not without some attention to detail.

If you are doing a sheet metal chassis, making a "model" out of posterboard can be very helpful. The posterboard is much less stiff than metal, but the general weak spots should be the same locations.

If you like doing FEA, you can apply moments in opposite directionon the center of the outer rails. Moments on the order of 180 to 240 in*lbs is a good place to start. Look at the displacements relative towhere the wheels would be. Displacements on the order of 1/16th of an inch or more are important. If you think about the "pile" of the carpet. A light touch to fully supported (60-75 lbs) is on the order of 1/8".

If someone was interested in doing some testing, you could anchor 3 corners of a frame onto something rigid (say very stiff table or multiple pieces of plywood). You could then apply weight to the 3rd corner and measure displacement . Trying different gussets, bracket, welding... and repeating the experiement would make for a very nice paper and possibly a cool sciencefair project.

[JesseK]

Originally Posted by IKE

... and repeating the experiement would make for a very nice paper and possibly a cool sciencefair project.

Personally I like to drive the full-weight robot at full speed into a brick wall while it has bumpers on. Driving it in at various angles so the corners hit also helps.

Why? Usually a frame stays square "just fine" until an impact. Then things can get bent out of shape.

[IKE]

Originally Posted by JesseK

Personally I like to drive the full-weight robot at full speed into a brick wall while it has bumpers on. Driving it in at various angles so the corners hit also helps.

Why? Usually a frame stays square "just fine" until an impact. Then things can get bent out of shape.

You might be missing the point, the frame is not "bent", but bends torsionally under load, and then returns to flat. When this occurs, diagonal corners touch down and diagonal corners lift. X=touching, 0=not touching:
X X O
O X X
This reaction effective turns somwhere between a long and a short wheelbase, and makes turinging more difficult.
In order to counteract this, teams will often add more "drop" to the center wheel. As the CG Height increases, this increases rock which can cause the center of rotation to oscillate between front of center wheels and behind of center wheels. The stiffness of the wheel/tires, carpet then act like a spring. The softer the spring, the more displacement, which in turn will equate to more energy (E=1/2KX^2 where as F=KX decrease K, and X increases proportionally in which energy increase due to the X^2 term).
This is why buck bobble is more prevalent in pneumatic tire chassis.

[Al Skierkiewicz]

Haochuan,
The "jumpiness" that teams describe is primarily the torque of your drive train overcoming friction with the floor. As a wheel breaks loose and the RPM jumps up, the result is a mini-wheelie when the tires again grab the carpet. In most cases with high friction tires (four or more wheels) this is the case when turning.

[FrankJ]

The extreme case is "help my robot will not turn" or once you go mechanum you never go back

In the real world: watch the skid steers on a construction site & how they chew up the ground.