Being from a team that lives at 7000 ft and dealing with bumpers/noodles shrinking when going down to compete in TX near sea level, the new materials should help alleviate that problem for us. Good stuff everyone. We’re just now thinking about what we want to try!
4 Likes
That is a fascinating detail that I know I’ve overlooked and wonder if it impacts teams who fly to events as well.
3 Likes
I am assuming your wheels are free-wheeling? Having them fixed might more accurately simulate a wheel being driven by a motor.
Dragging this back to bumpers 
@john3928 My colsons are tightened down with lag bolts and don’t rotate. I did find one rotating a bit after the first test run and re-tightened. I probably need to just go ahead and ruin them with an off center screw.
An important data item: my shock logger has a maximum of ±16g on a single axis and ±27gs for the three axis vector sum. We are fairly close to that. YMMV.
I tried to do a magnetic release to improve my data, but the door lock I had laying around turned out to be unable to pull the hammer back more than maybe 1/3 of the way.
On to the data! Test 1 is a flat to flat just as a checkout. Runs 2-11 are corner to flat. The corner on the hammer got wrap around hollow noodles.
I did not get -failure- on the corner noodles, but definitely some damage:
Collecting all the data together:
Here’s the knock back data; think of it as energy transfer to the sparring partner. I’ve got two reference steel blocks next to the front tires. I measure the distance moved back and average to report this.
1 Like
The lack of difference between corner and flat impact tests suggest to me that it doesn’t matter how much “noodle” is in the path of collision. The noodles aren’t doing much at all to reduce the impact forces in your test setup. I am interested to see how (if) that changes with other bumper materials.
3 Likes
Update to this - last event for this robot has concluded with no significant damage to the wood. I’ll cut them apart for a post-mortem in the next week or so. I think one of the studs may be working loose, but they’re all still secure, and none of the wood has split. By my math, we have played in over 100 matches on these bumpers.
1 Like
It looks like your floor is maybe concrete? Typically carpet would have a higher coefficient of friction which would transfer more energy into you test subject.
BTW, it looks like a 200g acceleration data logger could be built for well under $100 using ADXL375 - High G Accelerometer (+-200g) with I2C and SPI [STEMMA QT / Qwiic] : ID 5374 : Adafruit Industries, Unique & fun DIY electronics and kits and Adafruit Assembled Data Logging shield for Arduino : ID 1141 : Adafruit Industries, Unique & fun DIY electronics and kits and any Arduino-class device Arduino | Analog Devices ADXL343 Breakout Learning Guide | Adafruit Learning System
1 Like
Have you had positive experience using the logging shield?
I’m trying to build a similar data logger using the ADXL 375, 343, and 345 communicating over I2C to a Raspberry Pi with lackluster results.
Yes, its a concrete floor. One reason I picked Colsons is that they are pretty consistent over a wide range of surfaces. The ribbed treads are much grippier on carpet and get thrashed fast on concrete. What I’m doing is definitely going to make a flat spot!
CDers: I’m thinking about going to a lower impact energy to get more solidly in the range of my logger, but I’m worried that will be too gentle to be useful! Thoughts???
1 Like
The bumpers will behave differently depending on collision speed. The best performing foam in any given collision is the one that will compress fully as the robot(s) come to a stop. A foam that gives the lowest acceleration in a low speed collision will bottom out and give very poor protection in high speed collisions, and the foam that performs the best in high speed collisions will never fully compress in low speed collisions resulting in high acceleration than a soft foam. Because we probably want to minimize damage in general we want to reduce the peak forces which means focusing on the highest speed collisions we might expect. To me that seems like a 150lb robot (or analog) into an immovable target without padding at >20ft/s.
7 Likes
What @Weldingrod1 (or anyone else) can conceivably do:
What the community wants:
Lets be reasonable in our expectations people lol
15 Likes
Happy retirement to our 2024 bumpers. These lasted all season for us which was 60 matches total (34 blue) not including practice. The nested plate attachment to the frame worked well. All 3 DP angles that helped locate inside of the Mk4i cracked through. Triangular aluminum gussets all loose, should add tightening these to checklist and to use loctite. Wrap around pool noodles in corners cracked through as we all would expect. Wood screws holding wood together are clearly loose, but it all survived. Tongue and groove next year. Looking forward to building better quality ones next year.
2 Likes
No, I have not used that particular shield – and I’ve pretty much moved on to goofing around with the Orange Pi ecosystem (though it’s definitely not as inexpensive as the Arduino world).
I guess we’re going to have to reshoot with Barbie driving…
https://imgur.com/a/crash-test-1-2024-robot-slow-mo-1IEdK2W
Have you tried filling the hollow pool noodle with backer rod of a smaller diameter? The hole in the center of our noodles is about 3/4" and you should be able to find that size at your local hardware store.
You could definitely do that if you were desperate, but backer rod is about an order of magnitude higher quality than generic pool noodles, AND actually the right diameter.
As others have said, the size of the impact will affect the mechanics of how a particular foam/padding will perform. That said, I think there would be ways to scale the model that would preserve the important quantities. For example, halving the height of the bumper and the mass of the impactor should preserve the data. You could probably do some variation of speed/energy of the impactor and thickness of the foam, along with interpreting the acceleration and/or displacements to get some generic energy dissipation value, but that’s outside my current brainpower to work out, and also could break in cases where the structure matters, like a noodle around a corner.
I think we could treat either the hammer or target as the analog of the (to be damaged) robot. Measuring the hammer would give a very convenient way to scale the acceleration reading by just putting the sensor closer to the pivot. Being a more constrained object might also give cleaner data?
We should be able to crowdsource an experiment design that satisfies both reasonability and the theory nerds, if not those wanting an equal impact demonstration. I think that testing [an analog of] the worst case collision of two robots driving into each other full speed and ending at a stop, dissipating 2×½mv² of KE into the foam and no KE in the bots. Shifting reference frame on that collision to one robot stationary, one at double speed, and ending with both at the original speed (just after complete foam compression): we started with ½m4v², the robots ended with a combined 2×½mv², leaving the same 2×½mv² in the foam. Comparing those numbers to post 20 with 103J KE after impact + 37J (without internal rounding) left in the foam, the balance is pretty far off. I don’t know if that comes from a bad assumption in those theoretical impacts, the unequal masses of the hammer and target in the experiment, or weirdness/unideality of the noodles as padding (possibly including elastic effects?). A first pass estimation might be that the ideal padding would be twice as firm as those petal noodles?
Theoretically the CG is going ~14.2 ft/s with the given heights, but we don’t know what that translates to at the tip. A naive ideal would be to get the tip of the hammer to double robot top speed, which seems possibly feasible on the given rig? I’m not sure how important this is to the absorption dynamics though (damping and such). Obviously adding more energy would get it closer to match conditions, but I think it should be pretty safe to just halve the height (not the thickness) of the bumpers for the test, assuming the noodles don’t move and slip past each other instead of colliding directly. Half-height bumpers allows for half the masses/energies, though that’s still not quite match equivalent. Theoretically the flat/block foams could go even shorter, but not sure if that is worth it.
I’m not quite sure I agree with @Patrick3357’s suggestion of unpadded immovable target. That works for a purely symmetric collision, but I don’t think it would work for a corner-side impact. Similarly, a padded immovable target doesn’t seem to match that reference-frame-shifted real collision, something about acceleration during the collision? Immovable target does go with a single-max-speed (15-20 fps) hammer instead of a double-max (30-40 fps) one though, so I’m probably mixing mental models. Regardless, having the motion of the target as an additional measurement is useful.
If we assume that the hammer and target are staying mostly together during the skidding, and that the collision/absorption all occurs before the target moves appreciably, I’m not sure that the exact friction (wheels and carpet vs concrete) matters (so long as it is reasonably high). If we model it as initial hammer KE→some KE absorbed by foam and rest transferred to hammer+target→no additional KE released by foam and remaining KE dissipated as friction on the ground, I think our error would be capped at the foam displacement to move the target or remaining displacement after target moves vs the total target displacement? That feels a little sus though, and maybe I (or someone else check me) need to break it down a bit more and see how those compare more rigorously.
I think I saw some more detailed/knowledgeable posts on collision dynamics somewhere, but in some other thread it seems. I think we would care about either the peak point/load force for direct frame damage, or the total energy absorption during the collision as a more general take. Subjectively, the resilience of the construction and whether a frame corner will damage, wear out, or slip through the padding is also important.
@Weldingrod1, here are some of my immediate takeaways/suggestions:
- The accelerometer could be mounted partway along the hammer rod to reduce its absolute reading, then we multiply it back by the length ratio.
- If you make half-height bumpers, you should be able to get the equivalent of a real robot collision using half the mass/energy.
- Could you please post the dimensions of the hammer, specifically the pivot-CG/weights distance and the pivot-bumper distance?
- More generally, if you could give measured weights of the objects instead of estimations, it would feel more reliable.
- Also, if you can post the actual raw data and conditions of the impacts, rather than just graphs, we could maybe do some additional analysis. e.g. in post 20 it looks like you used 2 hollow petal noodles on the hammer flat and 2 solid backer rods on the target, but I’m not sure what specification of rod. I assume post 21 used the same petal-against-backer rod setup? In post 44, it looks like you had mixed round-petal hollow noodles on the hammer corner, was that also on the hammer flat? Was the target the same backer rods? The test 1 flat-flat control doesn’t seem to match the pattern in post 21, so maybe the conditions changed, maybe the data collection changed slightly, or maybe that was just a fluke and would’ve come out with a few more controls, but it’s hard to tell from only what’s posted.
- It might also be an interesting additional data point to see how much static foam compression it takes (just pushing on the resting hammer) to start moving/skidding the target.
Thanks! I hope others can chime in with their thoughts and check the validity of this post as well.
Super happy to get a really meaty reply!
I love the idea of the moving the sensor onto the hammer and moving upward to get into its operating sweet spot! I’ll try it!
I’ll have to come up with something to get official weights and report back. Bathroom scale might work.
Clearly, I need to post some high speed photos! I’ve ordered a quick release, so I should be able to do it soon.
On the impact side, the hammer and target really don’t move together much at all. You get the hammer mushing into the bumpers, then both of them rebound vigorously. The hammer swings back quite a lot.
I’ll take pictures and dimensions for stuff! The hammer arm is about 7’ long; it just barely fits in my garage, and the pivot is maybe 2" down from the ceiling. The hammer CG starts about 12" off the floor.
I’d prefer to keep the bumper construction exactly as a real robot, to avoid questions of scaling and accuracy. As we shrink things we will run into quite a few issues there.
I measured the force to move the target. I’ll have to see if there’s a good way to measure the foam compression to start moving. Or at least some way to get at foam spring constant.
Here’s the tabulated data:
1 Like
