Frictional Discrepancies

[Tristan Lall]

Originally Posted by AndyMark

Tread Material: White Acetal
...
Coefficient of Friction with plastic surface: 0.5 Static, 0.4 Dynamic

Originally Posted by The Kit of Parts, 10.2.4.1

The wheels supplied in the 2009 KOP are very different from previous years’ kit wheels. The tread material is Celcon M90, and has the following coefficients of friction on white, rippled fiberglass plastic sheet
Inline, static: 0.06
Inline, dynamic: 0.05
Transverse, static: 0.14
Transverse, dynamic: 0.10

There's an order of magnitude between the two sets of published figures (for the inline case, assuming that's how AndyMark tested them). This makes design difficult.

Any guesses as to which is accurate? I'd surmise that AndyMark tested early-production wheels, and that FIRST is quoting from a generic material specification, but that's entirely speculative.

I looked up a Celcon datasheet for bearing design, and (on page 67) they suggest that the dynamic coefficient of friction of Celcon on Nylon 66 is around 0.17, and that for Celcon on steel, it will be between 0.15 and 0.30, depending on speed and pressure. That's a third set of data (albeit only vaguely appropriate to our application), and it falls somewhere in between the other values.

Originally Posted by The Arena, 6.2.1

The REGOLITH is made of “Glasliner FRP^tm” gel-coated, fiberglass-reinforced, polymer material.

Also, note that Section 6 suggests that the floor is the gel-coated Glasliner FRP, while Section 10 (above) states that the sheet is "rippled". According to this and this, the "pebbled" and gel-coat finishes are mutually-exclusive options, so one section must be misleading. Which type of material was used to generate AndyMark's test numbers? Which material will appear at the competitions? (Apparently at the NH kickoff, it was the textured stuff.)

[Paul Copioli]

Tristan,

As far as the gel coat is concerned, both the pebbled and smooth are gel coated. the premium smooth finish has a thicker layer of gell coat. The gel coat is necessary so the fibers don't start poking out of the plastic after the molding. I am certain the pebbled finish is what we want.

[paulcd2000]

i would trust the manual numbers, because andy-mark is talking about "plastic" whereas the manual specifically refers to fiberglass. also, just by looking at the demo robots, i think mu is way lower than .5 otherwise, the robots wouldn't slid the way they did.

[StevenB]

Originally Posted by Tristan Lall

There's an order of magnitude between the two sets of published figures (for the inline case, assuming that's how AndyMark tested them). This makes design difficult.

Any guesses as to which is accurate? I'd surmise that AndyMark tested early-production wheels, and that FIRST is quoting from a generic material specification, but that's entirely speculative.

I think it's pretty clear that the AndyMark numbers are wrong. From what I remember of previous years with rubber on carpet, a mu of .6 or so was decent, and more than .9 was absolutely amazing. Mu of 0.5 would be only a small step down.
The numbers FIRST quoted sound like they were derived from testing, given that the manual has both inline and transverse coefficients.

[usbcd36]

It does make a huge difference, though. If the numbers from AndyMark are correct, it means inline > transverse, which makes skid steer a reasonable choice. If the numbers from FIRST are correct, it means the opposite.

[AndrewN]

Doing an incline plane test with four wheels and a frame we found almost no difference between the lateral and transverse static COF as measured by the angle of the incline. We are using the correct surface too.

Just to make it clear:

1. Point the locked wheels down the incline and gradually raise one end of the surface until the frame breaks free slides down. Measure the height at which the frame breaks free (8.5" over a 6ft sheet). This is a measure of the lateral static COF (tan of the angle between the horizontal and the incline). Our result is around 0.12 or an angle of 6.7 degrees.

2. Turn the frame 90 degrees. The wheels are now sideways down the slope. Repeat test. This is the transverse static COF.
Our result 8.5 - 9". Almost the same as the lateral value.

We expected that the heights of the two tests should almost be a factor of two or more different given 0.6 and 1.4 as the printed static COFs.

Can other teams please repeat this test and report the angles they are finding for both lateral and transverse static friction.

[martin417]

As an engineer, I have an issue with the "lateral" and "inline" numbers. On an ideal surface, and the interaction between the wheels and "regolith" this year com as close to ideal as you can get, friction depends ONLY on friction coefficient, and normal force. There is no directionality component. I looked at the wheels, and went to Home Depot and looked at the surface. I can see no reason, theoretical or otherwise, for a difference. As AndrewN's testing shows, in-line and lateral should be identical.

[Bongle]

Originally Posted by martin417

As an engineer, I have an issue with the "lateral" and "inline" numbers. On an ideal surface, and the interaction between the wheels and "regolith" this year com as close to ideal as you can get, friction depends ONLY on friction coefficient, and normal force. There is no directionality component. I looked at the wheels, and went to Home Depot and looked at the surface. I can see no reason, theoretical or otherwise, for a difference. As AndrewN's testing shows, in-line and lateral should be identical.

My team thought this was weird as well. Our theory is that since the wheels will mostly be spinning forward and back more than they'll be sliding sideways (ignoring any holonomic-type drives), then they'll tend to get heavily scuffed in a forward-back direction. So perhaps after a few matches worth of use, you might see the kind of transverse coefficients that the manual describes.

[rfolea]

Originally Posted by AndrewN

Doing an incline plane test with four wheels and a

Can other teams please repeat this test and report the angles they are finding for both lateral and transverse static friction.

We measure COF by drag - we simply drag the load with a scale attached on a flat surface.

We are getting around .1 in either direction, dynamic. Static was sightly more (.12 I think).

There is no noticeable difference between dragging it sideways or not with the wheels locked.

[Matt C]

My only guess is the testing may have been done on the wheels with the mold lines still on them (not worn)?

Could this be trying to dig into the field material, thus raising the effective CoF?

[writchie]

Originally Posted by AndrewN

Doing an incline plane test with four wheels and a frame we found almost no difference between the lateral and transverse static COF as measured by the angle of the incline. We are using the correct surface too.

Just to make it clear:

1. Point the locked wheels down the incline and gradually raise one end of the surface until the frame breaks free slides down. Measure the height at which the frame breaks free (8.5" over a 6ft sheet). This is a measure of the lateral static COF (tan of the angle between the horizontal and the incline). Our result is around 0.12 or an angle of 6.7 degrees.

2. Turn the frame 90 degrees. The wheels are now sideways down the slope. Repeat test. This is the transverse static COF.
Our result 8.5 - 9". Almost the same as the lateral value.

We expected that the heights of the two tests should almost be a factor of two or more different given 0.6 and 1.4 as the printed static COFs.

Can other teams please repeat this test and report the angles they are finding for both lateral and transverse static friction.

Perhaps you have discovered what's behind the "fish" clue

The transverse/inline ratio is a very critical parameter. If it's no where near the 2.3 advertised, then many preliminary design decisions about drive configuration will be dead wrong. We will try to confirm your findings as soon as we can locate the actual surface material.

It is possible that the ratio changes significantly with normal forces closer to 1/4 of the nominal weight of the robot due to the way the materials deform under load. It could also be that the type of backing underneath the regolith is a factor. The wheels are very hard and provide a very small contact area. If the backing is carpet (rather than a very hard material), there could be a small depression that presents differently in transverse and longitudinal directions. Your numbers may reflect light loading, before such effects manifest themselves. Based on your data it does looks like we will need to confirm the Mu values under a range of loads.

Does anyone know whether the regolith is over carpet?

Good catch.

[SWIM]

Originally Posted by writchie

Perhaps you have discovered what's behind the "fish" clue

The transverse/inline ratio is a very critical parameter. If it's no where near the 2.3 advertised, then many preliminary design decisions about drive configuration will be dead wrong. We will try to confirm your findings as soon as we can locate the actual surface material.

It is possible that the ratio changes significantly with normal forces closer to 1/4 of the nominal weight of the robot due to the way the materials deform under load. It could also be that the type of backing underneath the regolith is a factor. The wheels are very hard and provide a very small contact area. If the backing is carpet (rather than a very hard material), there could be a small depression that presents differently in transverse and longitudinal directions. Your numbers may reflect light loading, before such effects manifest themselves. Based on your data it does looks like we will need to confirm the Mu values under a range of loads.

Does anyone know whether the regolith is over carpet?

Good catch.

From how I interpreted the rules, the entire field is covered in carpet, and the regolith is placed over that.

[Ozeaden]

One thing that my team found out is that if u wear down the wheels, you get better traction. With having a rough tire, it will give more of a stick to the flooring material. We tried it on one of our past robots and it worked really good. Just run the wheels on asphalt and run it down a bit. Its not against the rules at all.

[comphappy]

Originally Posted by Ozeaden

One thing that my team found out is that if u wear down the wheels, you get better traction. With having a rough tire, it will give more of a stick to the flooring material. We tried it on one of our past robots and it worked really good. Just run the wheels on asphalt and run it down a bit. Its not against the rules at all.

Read <R06> and I think you will find you are very wrong, and are now out $100 in usable wheels. This is intentional damage, which is explicitly prohibited by that rule.

[SWIM]

Originally Posted by comphappy

Read <R06> and I think you will find you are very wrong, and are now out $100 in usable wheels. This is intentional damage, which is explicitly prohibited by that rule.

I'm sure they'll get away with it, unless they're worn down to the point where it's obvious they were doing something far beyond just driving the 'bot around. But really, getting grooves in the tread shouldn't change the coefficients of friction much at all, since the surfaces are so hard.

If they do try and cheat to get more traction out of their wheels, that's a pretty rotten thing to do, but I don't think other teams will have to be too concerned with them somehow getting considerably more traction with worn down tread.