Hello,
I'm facing a what-am-I-looking-for problem again. I've got tracking data, meaning I tracked myself riding a bike a couple of times on the same track in order to have some test data. So I've got time-distance pairs. I want even more different ones, but want to generate them. I want the virtual test drivers to be both faster and slower. I don't want it to be a linear summing up algorithm.
Basically I just want a hint, what I could search for. Can somebody help out?
You could use a pseudo random number generator to generate random variations based off your original data. Just provide a range +/- your measured values you'd allow the randomness to go, and generate random dataset values for time + distance. This would let you generate as many test cases as you need.
Try multiplying each data point's time value by a constant. If it's less than one you'll be creating a faster test drive, and if it's more than one you have a slower test drive. You can also fudge the position similarly if you want.
You could check out Markov Chains for generating random data although it depends on how complex you want your distribution of data to be. For a simpler approach, Reed Copsey's solution is going to be easier to implement.