First of all, I apologize if this is not the right place for this question, I wasn't sure whether it would be more appropriate to post it on stackoverflow or here. I should mention that I am not really knowledgeable when it comes to low-discrepancy sequences.
I am writing a program that relies on using a low-discrepancy sequence for sampling in 2D. Hammersley, Sobol, blue noise, and the golden set work just fine for the problem in general. My issue is that I need to have m sets of n points, and each of these sets should be low-discrepancy, the union of these sets should also be low discrepancy. My question is whether this is feasible at all, and whether anybody knows of a paper on the subject.
you want e.g. to take two 2D point sets to form a 4D set? This will usually not work fine, and of course there are algorithms to directly generate a 4-dimensional low-discrepancy set (or 8D or whatever, like Sobol). It is most easy to see from a 1D analogon: Taking two 1D equidistant sets leads to a 2D set looking like a 45-degree stripe, having very bad space filling properties. Maybe it is good enough for you to take two 2D set with some randomization, so that at least e.g. the combination of 2nd and 3rd variable has random properties (but not low discrepancy)??