One way to bias the docking search towards the ‘xy’ plane would be to use constraints (.cst file). Specify residues along the ‘equator’ of each of the proteins to have be favored at the interface and residues at the ‘poles’ (z-axis) of each protein to be disfavored at the interface. That way the docking search will be biased towards complexes that have ‘equator’ residues at the interface, enriching xy sampling.
You’ll have to look into the number of residue constraints needed. Depending on the size of the proteins, I’d guess around the ball park of 5-10 equatorial residues on each partner and 2-4 polar residues on each partner.
As for the original question, Aroop is correct. Standard docking was not designed to limit the docking search in this manner.