Few more thoughts on velocity obstacles.
Firstly, I'm also fed up with performance figures provided in FPS figures. Humus had a good post about it from graphics programmers point of view. The ClearPath paper uses frames per second all over the place to describe the performance of their algorithm. Academia, if anything should embrace the use of the System of International Units, and their peer review process should enforce to provide more information how the performance was measured, just like they require the author to state the prior art too.
Secondly, I forgot to mention another recent VO paper:
Independent Navigation of Multiple Robots with Hybrid Reciprocal Velocity Obstacles
The paper provides nice trick to avoid the reciprocal dance problem usually related to VOs.
So, now there's VOs, RVOs, NLVOs, FVOs, HRVOs, NHRVOs, PVOs and what what not. And with every new VO related abbreviation there is one less reason not to invest in proper crowd control on your game.