New VarTree24/48 Storage Technique For Position Encoder
EliCDavis opened this issue · 0 comments
EliCDavis commented
Instead of evenly breaking up space to discretize points, let's instead subdivide space in each dimension by a variable amount. The number of bits dedicated to a given dimension is based on the size ratio of that dimension when compared to the others. So if the total range of movement in the X dimension is half that of Z and Y, then X will only have 4 bits dedicated to it instead of 8 (for the 24-bit version).
This method would get us 2D for free, and overall just boost the accuracy of recordings.