Sorry for my previous answer, I look at internet and Petrel has Multipoint Geostatists, to better understand it I recommended to read the works of Stanford University and the library GSlib
If people are still wondering Petrel (at least the new one) does indeed have a range of geostatistical modelling techniques including one for MPS. RMS will also run it. Definitely recommend most of the stuff that comes out of the Stanford Petroleum Engineering department. The only problem is that the literature does actually tell you how to model (in Petrel) only the statistical concepts behind the technique and their various modelling algorithm. Also as far as I am aware Petrel will generate the TIs and run MPS for you but without telling you exactly how it achieves the end result. Has anyone else encountered this and/or know a way around it?