I have three sets of data that were sampled at the same interval (20hz). The data is from a data logger that measured vehicle speed at a specific time.

In each of the tests, the Car starts at the same starting point A, and is manually drive to the end point B. This was performed 3 times. The point at which the vehicle comes to a stop is different each time since it is manually driven but that is intended.

My intention is to gather enough data sets from this test route so if another person drives the test route, we can compare how he performed.

I would ultimately like to average the data sets from my run and call it the optimal path and add an error band, so if someone else runs the test route, I can compare how they did to it, like distance from end of route, and max speed through a turn.

Initially I thought I could plot the data vs distance and take the average that way but the problem I am running into is, log A is 400 samples, log B is 420 samples, log C is 440 samples. So they arent exactly 1:1. Any idea how to skin this cat?


1 Answer 1


Here is one idea.

First plot the time vs distance. These lines would have different slopes between the measurement periods and you would essentially estimate the speed between measurements. Here is a graph with one measurement window:

enter image description here

Since they're racing on the same track, the total distance driven should be approximately equal. You might need to stretch or shrink the height of lines to make sure they travel the same distance. (This might not be a good assumption for your case.) Now, go from the y-axis and for regular distance intervals, look up the slope of the graph. Then plot the observed estimated speed for each distance period. You'll end up with speed measurements at regular distances which will now have the same length of samples.

This way you'll have a fairly good approximation of the speed at which racers traveled at specific parts of the track. I.e. you can compare cornering speed etc.

  • $\begingroup$ I believe I ended up doing something similar albeit less calculated. I wanted distance from 0-100 meters at quarter meter intervals so I pulled that from the car's wheel encoder. This allowed me to plot speed and acceleration with respect to a path. I was then able to line up my data points from each log and compare things like cornering speed, brake initiation, and stopping distance. I really like what you did. Can I apply a similar technique to other metrics I might be interested in? $\endgroup$
    – ChipsAhoy
    Commented Mar 22, 2017 at 18:19
  • $\begingroup$ I assume you could. In general, the method I outlined above helps you "pivot" your data from a non-aligned measurement axis (the time intervals) to a new axis that has the same lengths across the samples. $\endgroup$
    – niczky12
    Commented Mar 23, 2017 at 8:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.