Dear LB users,
my code currently implements Zou and He pressure and velocity boundary condition schemes for either 2 or 3D simulations.
I would like to do an “upgrade” introducing the Regularized-like implementation of velocity BC for the D3Q19 model.
The process explained in the review paper by Latt and others (Straight velocity boundaries in LBM) seems to be quite clear about that (Section C., Regularized method (BC3)), except for edges and corner nodes, where there are always one or more couple of opposite lattice directions which are both unknown, since the corresponding distributions are streamed from outside the computational domain. Therefore it is not possible to bounce back the non-equilibrium part of these distributions as described in the paper for the preliminary calculation of tensor PI(1).
Therefore my question is how to implement regularized method (BC3 as referred to in that paper) on these kinds of nodes?
Thanks to you all and many special thanks to whoever will like to answer or to give me any kind of help 
Ernesto