[SamPlay]

I agree that there is no absolute solution. I just meant that compared to what I have seen on PVS computation( portal based ), the solution of J Bittner seems more efficient, and more robust( but I am not an expert!).

My interest is not in PVS computations, I am presently starting to think about whether an automatical map tactical analyzer is feasible. This means in particular knowing behind which face a bot can hide from potential enemies in neighbouring "regions". This is basically a problem of visibility/occlusion and that is why I looked first at PVS computations methods.

Besides, working with 4D is a way to address move issues of rotating hulls, by adding yaw( heading) to the 3 cartesian coordinates.

As regards Bsptree, I agree that the algorithm works in any dimension, but the splitting of a N-dim polyhedron is a problem to me.
Sort of going back to 3D does not work, as using Plucker representation of lines in 5D is the key ( and smart) thing in J Bittner solution.
Thanks for your comments!
NB: the title of Jiri Bittner's thesis is "Hierarchical Techniques for visibility computations"