RPM(reflecting particle method)

Hi all,
To induce the steady flow(NEMD), there are mainly three method, EF(external force), DCV-GCMD(dual control volume-GCMD) and the so-called RPM(reflecting particle method). And I have run the former two in lammps. But come to RPM in lammps, I found it seems difficult to run in lammps. The most important step is to creating the fictitious membrane. Paiticles can be reflected in a probability. In contrast, particles can flow through the membrane free from another direction. The article https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.102.184502 have described the method in detail. I think the method is the best way to induce the flow. But, unfortunately, I found there are no fix can realize it. And the fix wall/reflect seems similar. In fact, the wall/reflect fix reflect the particle in 100% probability and particle can not be able to through the wall from another direction. That is so different with the RPM. So, could you give me some suggestions about RPM in lammps? Anyone have run the method in lammps? Thanks in advance.
Dongbo Wang