6WD Chassis Stiffness vs. Maneuverability

For the last 3 years KB33 has been build 6WD aluminum sheet metal chassis. It has been an iterative design mostly focusing on more robust drivetrain. One thing we think we have noticed is that some machines seem to be more maneuverable than ours. In 2006 we were underpowered and had a difficult time executing turns. In 2007, we did a good job of holding our own, but there were faster and more agile robots. For 2008, we ended up going with an asymmetric chassis set up with 3 omni-wheels on three corners to improve turning. This was OK for Overdrive where D was mostly played on the ball, but would have been death in 2007.
So here is the question for 6WD teams:
Have any of you noticed chassis stiffness (most likely torsional) making a big difference in agility? I know that rock (amount of lift or drop on 1 set of wheels) plays a huge part, but with similar rock, we just don’t seem as maneuverable as 1114, 254, and several others… Could some of the iterative 6WD teams please chime in on this. If it is trade secrets, I understand, but if you care to share your knowledge I have open ears. I did a lot of analysis on slip and traction curves vs. speed and power consumption. My calculations model our chassis very well, but I was amazed to see that certain teams seem to defy the physics I know. Is it an illusion and you guys struggle with scrub balance, or do you just have a better tuned chassis?

P.S. Let’s please stick with 6WD design discussion. I realize there are a lot of pros and cons to other designs (swerve, halonomic…), but would like to keep this 6WD related.

Maneuverability - i.e. turning rate, is base on several things.

Torque (inversely proportional to speed), especially if you are skid-steering.

Friction, i.e. wheel material type.

Weight distribution front to back.

Wheel base, both in the x and y dimensions. “Rocking” your wheels, or lowering the center one, is a way of giving yourself a wide wheelbase because you generally only have 4 wheels that are primary load balancing wheels.

If you are looking at ways to improve your turning, ask yourself this:
Assuming that your torque is fixed,

  1. Do we need friction on the front and rear wheels? Some people add zip ties around the wheels to lower it. Some people “flip” their treads. Some people use omnis in the front and rear. Some people use wedgetop rather than roughtop.

  2. Is our weight distributed in the center of the robot? If you are weighted heavily to the front or back, it will affect how you turn. The easiest way to experiement is try with your battery in different locations and see how the robot reacts.

It’s my guess that they either have more rock than you do, a different weight distribution, or less friction on their front and rear wheels.

Editted to add: I am by no means an expert on this - we’ve fought our own battle this year. We want straight line stability for autonomous, but it doesn’t lend itself to easy turning. We went through a number of iterations this year with our drivetrain trying to find something that worked well. In the end we had to gear down more than we like just to be able to turn, and our top end speed suffered a bit as a result. We never did get that 5th auton line.

In my experience it is VERY possible to do a 6WD with 6 high traction wheels. If you find yourself reducing traction on some wheels to increase turning, you’re negating MANY of the virtues of using a 6WD in the first place.

There have been lots of arguments about optimizing a 6WD configuration, especially centered on how much “wheel drop” you should use. In my experience this isn’t as huge a factor as many people seem to think.

I believe that the key to a well performing 6WD lies in the robot’s CG. A lower CG will result in much better performance than a higher CG. Placement front-to-back is also important and will greatly change the way a drivetrain performs.

Don’t believe me? Take one of your old robots and add some weights to it in different places, observe the change.

Good Luck! Looking forward to seeing Team 33 at IRI.
-John

1 Like

We have had this problem as well in the two years we’ve done a 6WD. Even in '08, when the front wheels were omni and and two rear traction wheels were spaced only 12" from each other, we still had issues ('07 was a rocker). Come to think of it, our ramps were mounted on the outside of the frame in '07 and the mass of our elevator in '08 came down right on the inside of each wheel.

This makes me wonder if the mass distribution left to right on the bot should look fairly centric rather than outward.

I think John has it exactly right. The lower CG machines - and the ones mentioned were lower CG than the 33 machine - were able to make the turns in a smooth and fast way.

I don’t know this for sure - it would be interesting to get some comments from the 1114 or 254 people - but did they bias their 2008 fore-aft CG so that the bot is running on the back wheels through the turns? My guess is that they did.

One other item that you mentioned in the title is chassis stiffness. I have always thought that 33 machines have had a somewhat flexible chassis. Maybe this is causing an unintended wheel contact through the turns as the chassis flexes?

The previous 33 design, with short wheelbase center+rear drive (like you had in 2002 I think) was amazingly fast and maneuverable. I was surprised you guys didn’t use that layout for 2008.

Ken

We haven’t had any problems with 6WD drive for speed and turning ability. We used 6 4x2 IFI traction wheels with the center wheels have a 5/32 in drop which was just enough to allow us to turn well yet retaining pushing power as well. All six wheels had the gum rubber thread from mcmaster so it really stuck to the carpet well. I think the key here is what john mentioned earlier is your robot CG!!!, we had 60% off our total weight four inches off the ground which made a great difference in our mobility. If done right you don’t need any omni wheels in the front or back to turn smoothly.

254’s 2008 Robot had an extremely low center of gravity. Also, the vast majority of our weight was in the rear of the robot mainly due to the following things:

  • All elevator mounting points were within 6" of rear wheel axle
  • Battery positioned slightly to the rear of center axle
  • Compressor (and most pneumatics) were on rear of robot
  • Due to open front of frame, all electronics (except for speed controllers) were mounted rear of the center axle
    I believe that our center wheels were dropped 3/16", and all wheels are 4" diameter and 1" wide, using black roughtop tread

Due to the weight distribution on our robot, the robot rarely had any weight whatsoever on the front wheels. When changing tread throughout the season, the front wheels rarely required a change (they probably could’ve gone the whole season on one set of tread). However, the center and rear wheels burned through tread much quicker (during competition, we typically changed the tread about once per day of competition, but always had a spare set of treaded wheels ready)

During play, the robot was probably on the rear/center wheels over 90% of the time, and only really rocked during quick stops and reversals of direction.

Thanks. I really appreciate the great feedback so far.

Here are some observations of our chassis developement. There is a bit of a monkey wrench in the “CG” height thing.

During the build season, we took it for a test drive when it was still a bare chassis and balasted with sand bags up to competition weight. We used the sand bags to change for-aft weight distribution, but the “CG” height could not have been more than 7" off the floor. The same dynamic issues that we see in the fully prepped machine were there with the bare chassis.
The omnis were added to improve cornering. When we had omnis on all 4 corners the chassis predictably had a lot of whip. If the “CG” was not balanced for aft, it was extremely loose (unstable). When I ran the turning delta power calculations it was under 50 Watts (delta side to side) for a relatively high speed tight turn. This can be achieved by just the slightest release of inboard power with a 2 chip per side powertrain. When we had 2 KOP wheels on the back the chassis required a lot of reverse inside torque to turn even at high speeds (approximately 300 watts calculated for the same turn that was 50 watts above). This was correlated with motor temps as example 1 the inner motors stayed cool, while the motors were significantly hotter in example 2. For competition we went with 1 omni (inside rear) and 1 KOP outside rear, high traction wheels in the center and omnis up front. The calcualted turning power (again for similar turns) was about 150 watts. This set up has been good for our machine for this year, but would lack in 2007 games.
As JVN stated, I know it is possible to run with 6 high traction wheels with relatively little rock (I have seen these machines and talked to them about their “rock” or “drop”). It just isn’t “working” for us. That is why we were thinking stiffness plays a big role.
(“Working” is relative term here in that we were able to repeatably get 5 and sometimes 6 lines in Hybrid on clean runs, so it does handle pretty well. I am really looking for turning good into great refinement.)

Ken,
The 2 traction 4 omni set up performed similar to what the 2002 did (so I am told, I have only 4 years with the Bees). With the high speeds of Overdrive it was just a little too unstable.

Tom- Talk to Jim Z. and he can give you some really cool tips on how to drive straight. With a gyro and some tuning he can help with the straight equation.

The cg idea is interesting. It fits our 6wd robots; they were both very maneuverable and had a very low center of gravity.

The frame twisting is also interesting… our 2007 robot had a ton of flexibility, twisting wise, in the frame. If you grabbed the arm mast and move it to either side, you could make the outside wheels (on opposite corners) touch the ground simultaneously. It never happened when we were driving around though. (the arm only weighed 15 lbs after all)

It may be possible that your problem is a combination of relatively high cg and a twisty frame. If you try to take a corner with all of that weight higher up it may twist your frame enough to effectively eliminate your dropped center wheels. Just an idea. (maybe someone clever could figure out how to test it…)

EDIT: hmm, your sandbag test is interesting. How high up were the sandbags, and how were they mounted? Our 2007 chassis was only six inches high, and only the arm extended above that amount. The arm was also mounted in the middle of the back end of the chassis (no braces to the corners), maybe that has something to do with it?

The fact that you are calculating turning power requirements is a cool thing. We never go into that level of detail. Turning is one of those things that we do experimentally (for better or for worse) and based on prior experience. (Guess and Check is probably a bad way to do it, but it seems to be working out so far. Negative reinforcement of the worst kind.)

What motors were you running in your drive this year?

-John

A low CG certainly does help you withstand higher turning moments, but it does not help you generate them.

For those that have read Chris Hibner’s whitepaper on turning (http://www.chiefdelphi.com/media/papers/1443), which is everybody I hope, it looks at the individual contribution of each wheel to the turning moment, so I suggest we start here. The moment is generated at each wheel by the coefficient of friction (CoF) and normal force (Fn) at each wheel. I think that the CoF is largely understood as demonstrated by the unusual wheel choices cited above. So what changes the Fn at each wheel? Simplifing the frame as a series of beams for discussion we can reference some stock solutions for the beams with 3 supports; based upon a uniform load the 3 moment equations show that the center axle would support 62.5% of the load with 18.75% for the outer axles. A snapshot of this eqn is in a previous post here.

JVN:
4 cims into slightly modified AndyMark gen1 shifters set up for 7 fps and 14fps (not sure on top end, I have slept too many times since then). We were going to add a FP to the outside drivetrain but needed it for the elevator. Jim Z’s kids are actually the chassis crew, and mine were responsible for the elevator. I did the chassis turning power calculations to see if we were missing something in our optimization process. Our first wheel selection had the driver full power on the outside stick and partial revers power only on the inside side. I did a lot of Dynamics calculations because I thought Overdrive would be even faster than it was. I thought that robots like 148 might be capable of up to 20 laps with a clean track. I was going to post some of the power calculation stuff until i saw 1114 driving around even faster with 6 grippy wheels.

s_forbes:
The sandbags were laying on top of frame members that are 5 inches from the floor, and were duct taped in place (got to use duct tape during build season since it isn’t allowed later). They are flat 25lb bags that when flat are only 1 inch thick. A ballasted chassis can tell you wahat Vmax and acceleration rates are for your robot. If you have poor theoretical top speed to actual top speed this can be a symptom of drivetrain efficiency issues (mis-aligned chain, or bound gear-sets).

33 CG height:
As far as CG height goes, our CG is actually about the same height as 254. 33 looks deceptively top-heavy due to the poly-panels. We do get a very high CG when the elevator goes up, but otherwise about 70% of our weight is below 7" and another 15% below 1foot line.

Thanks for the white-paper. These are pretty much the identical equations I used in my turning power spreadsheet. My assumption was that with rock I would be a 14"x28" Wheel base by track robot. I then used different mu values for longitudal and lateral values based off of experiemental day from the AM website and some of the good folks from 494. If you multiply the skid force by the velocity you get Power. Just make sure you use velocity and not simply speed as the direction component is very important to the overall amplitude.

This past year we went with 2WD but in 2007 we had 6WD. I found the robot that year to be very maneuverable and we only had, I believe, a 16th inch drop on the center wheel along with six two inch wide traction wheels. I would be inclined to agree with you that chassis stiffness is a big factor in turning ability because that year we had a welded box aluminum frame that was very stiff. Testing may be the only way to get very good answers. If you have the time to spare try building a few different chassis designs before the build season and test out their turning capabilities.

Apparently that got submitted early… whoops. To finish my point…

<theory>
When driving on concrete (or a similar surface) my observations have been that the chassis drives very similarly to how this distribution suggests, however, when we put them on carpet it can be very different. Many attribute this to the CoF change on the carpet, but I think that the Fn also changes as the carpet compresses under the bot like a spring. (See attached spreadsheet.) This is why lowering the center wheels improves the turning of a chassis. Yes, the center wheels can be lowered to the point of rocking, but my 3+ minutes of bot driving time suggest that rocking between 4x4 cg aft and 4x4 forward is not a pleasant experience. A lot of folk do manage it though and they are amazing.

So how does the stiffness of the frame play into this…

Assuming the carpet is spring ‘like’, the deflection of the frame limits the difference in deflection of the ‘springs’ under each wheel which eats into the wheel drop. The stiffer the frame, the less wheel drop lost.

In working terms, a stiffer frame requires less wheel drop to move the weight to the center than a less stiff frame. Consequentially a stiffer frame is also more greatly influenced by a change in wheel drop than a less stiff frame. i.e. a more flexible frame is more robust to wheel drop.

</theory>

Unfortunately, the situation is statically indeterminate so I don’t have a set of equations to conduct an optimization study with.

20080703 - Carpet Compression.xls (26.5 KB)


20080703 - Carpet Compression.xls (26.5 KB)

to improve maneuverablilty/decrease turning radius, our team [1002] has moved away from standard wheel setups and gone with 6WD, but with the middle two wheels offset by 1/16th of an inch from the other four wheels

this way at any given time, we have 4 wheels on the ground… usually… and one set of wheels 1/8th inch above the ground

you might not get the best results with this however though [as i said before] it does improve turning radius [and power expenditure for turns] considerably

I’m not too good with numbers, but I’ve built a lot of prototypes…

I’ve discovered one important thing. The closer the traction wheels are oriented into a square the better the robot turns.

This is why 4 WD robots in the “long” configuration turn horribly. Because the wheels are in more of a rectangle than a square.

Many teams have solved this problem by adding another set of wheels in the center to essentially make their robot into two 4WD trains. As the robot tips back and forth the wheels remain in a square (all six are never touching the ground). I believe this is why 254 is so speedy.

Another way to correct this problem is to place two omni wheels on one end of the robot. Since the omni wheels offer no side-friction, the robot will turn just like a 4WD machine with all four wheels in the back of the robot.

We’ve always built robots with the rocking syle drive train… and then gone back latter and added omni wheels :slight_smile:

I totally agree with JVN, ect. that say COG is the most important factor to a smooth driving 6-wheel drive. just try driving only your base drive system, it will drive like a dream, once you lump stuff on top and raise the COG it will never drive the same. comparing your robot this year to 1114, 254, etc. your COG is probably a good amount higher because of your scoring mechanism, thus leading to a tippier rougher driving robot.

Looks like we will have to run an experiment. I will do a tilt test to estimate actual CG height.

I really think that the poster talking about carpet stiffness vs. chassis stiffness is on to something. I think we will build an experimentally stiffened chassis.

NICKE/254? You said that you had to retread wheels daily on your middle and rear wheels. Was the tread-wear even? More on the inside? More on the outside? I do a little amateur racing and you can tell a lot from your tires…

In 2006, when I was still with 1293, we used the kitbot–stock ratios and all–feeding one IFI wheel in the center and Skyways on the corners. The videos I’ve seen all show us whipping around pretty hard, and we were quick enough to do our job (defense) effectively. It didn’t get us into eliminations, but it was good enough to be the best finish the team has ever achieved in qualification rounds (before or since).

In 2007, I took the concept with me to 1618. This time, we used skinned-and-roughtopped AndyMark FIRST wheels, driven by a Gen2 AM Shifter using the big and small CIMs. We used the second speed to blitz across the field, important when stopping folks on the far side of the rack. Observation from off-season testing and Brunswick Eruption showed that this design was a little harder to turn. It doesn’t like to turn in place, though it’ll do it when you push it hard enough (as I found when one of the pre-rookies on 2458 about spun the numbers off of it at Brunswick Eruption).

This year, we switched to the AM Super Shifter with two CIMs per side for lack of the large CIMs. (And, truthfully, we would only have used other motors on drive if we really needed a CIM elsewhere.) Instead of roughtop, we used wedgetop cut with the long diamonds going around the wheel (as opposed to how IFI cuts theirs) for a little less CoF since we were flirting with the upper limits of current draw if we stalled a motor. (That was before we never got to a manipulator and weighed in south of 75 pounds.)

The more revolutionary change for us was riveting the frame together instead of using bolts. The result was the tightest frame I’ve ever had a hand in (well, excluding Bob in 2004 where we sent out the 2x4 aluminum of that year’s kit to get welded together). The weight was slightly skewed towards the rear, with virtually nothing more than a foot off the ground. When driving both, I’ve found Speedy Debris to be a little easier to turn than Uppercut before it, though it may be a function of the 30-pound weight difference. An apples-to-apples comparison of bolted frames against riveted frames would be interesting.