paper: 5406 Acceleration Model Spreadsheet

Thread created automatically to discuss a document in CD-Media.

5406 Acceleration Model Spreadsheet
by: nuclearnerd

This is a re-implementation of Ether’s acceleration model in Excel format . The spreadsheet allows quick tweaking of parameters to compare with measured data. Four sets of test data are included.

As discussed here: https://www.chiefdelphi.com/forums/showpost.php?p=1715960&postcount=29, I have attempted to match Ether’s acceleration model https://www.chiefdelphi.com/media/papers/2868 with recent data measured by team 234 and our own tests at 5406. In order to speed up the process, I re-implemented the C code into a spreadsheet (no macros required) which is attached.

The spreadsheet includes four sets of test data:

  1. An output table from Ether’s simulation, to check that the two codes give the same result
  2. 4 cim high gear acceleration data from 234’s tests
  3. 6 cim high gear acceleration data from 234’s tests
  4. 6 cim low gear acceleration data from 5406’s tests

The code matches test data 1 exactly for the same input parameters, indicating no bugs were introduced when transferring formats. I was unable to get a consistent set of parameters to match all three remaining test data sets. This means the mechanical and electrical resistances were significantly different between data sets, or I’ve misinterpreted some other parameter. I was also completely unable to get the current draw to match the 234 data, even though I could match the speeds and distances pretty well.

Feel free to play with the parameters and see if you can get a better fit. (And please post them if you do).

The yellow cells are unknown tweaking parameters. The beige cells are known parameters that should match the tested robot without tweaking. The orange cells are for future use, and currently do nothing.

Interestingly, after playing with these for a few hours, the six yellow parameters don’t appear to be completely independent - you can get a similar result with a low Kf and low Kro as with a high Kf and high Kro for instance.

The goal of this exercise was to build a model that could be used to make informed choices about drivetrain gearbox sizing in 2018. (how many motors to use, what is the effect of different ratios on sprint times, when should we shift etc). While the model didn’t yield a “universal” set of parameters, it does give a good sense of the usual range for each parameter. For 2018, we will find parameters that under-predict the available test data and use that for a conservative estimate of acceleration.

Future work will involve adding voltage and current limiting and voltage ramping into the model, shifting at a prescribed time, and perhaps deceleration. License is CC0, public domain

5406 accel model.xlsx (19.1 MB)



Many of the parameters in the yellow boxes can be directly measured using methods detailed in our drive characterization paper. It would be interesting to see the model fit when using empirically-measured parameters.

The “orange boxes” you’ve added are very important, as well; clearly, something has to be done to account for the initial “ramp up.” This will be present even if no voltage ramp is used, due to inductance in the system.

Definitely need to take some measurements. The electrical resistances are easier to measure that the mechanical resistances, but your paper is helpful thanks.

Good point. I’m still able to match the data pretty well without voltage ramping, either because it happens so fast it doesn’t affect the later results much, or because I’m over-fitting. It will be good to get that in the simulation to check. Unfortunately I doubt I’ve got time before kickoff to do it.