Academic Facets: News and Views from Spatial's University Community

[#1]

And so to San Francisco and the “Geometric and Physical Modeling” Conference (GPM09). This annual event started off in 1991 as the “Solid Modelling” conference and although it has broadened its focus over the years it is still a natural home for a lot of CAD/CAM research that is nearer to mathematics and computer science than engineering.

Sure, there are other conferences (e.g. ASME/IDETC) where you could find an audience to discuss hard core academic work on topics like blending, feature recognition or Boolean algorithms, but few have the cachet of GPM. This is not least because of its super-harsh reviewing process. To be accepted every paper has to “survive” five separate reviews and unlike a journal there are no second chances, paper are either accepted or rejected; so no years of R&R (”Revise and Resubmit”) that authors sometimes endure before their masterpieces appear in print. Consequently one can be pretty certain that every paper which makes it into the GPM proceedings is worth reading.

Without dwelling on individual papers I can offer some overall impressions:

Firstly GPU implementations continue to steal the show; not just fast demonstrations of, algorithms that, for want of a better phrase, are “standard” problems (see the paper “Accelerating geometric queries using the GPU” ) but demonstrations of radical ideas that not so long ago would have been easy to dismiss as incomputable.   A good example of this trend was the theoretically/mathematically beautiful representation described in the paper “Configuration products in geometric modeling”, where the authors were able to report a GPU prototype of a pretty abstract idea (for me, at least, more than three dimensions is hard to visualize).

Despite the conference’s often rarefied academic atmosphere there were still examples of “engineers” using cunning to solve difficult problems in simple and effective ways. In this vain I enjoyed the paper  on ”Testing an axis of rotation for 3D workpiece draining”, because it didn’t employ CFD but a notation of gravity acting on a mesh surface to roll a “virtual” liquid drop across the network of vertices and edges. An example of a good observation beating brute force hands down.  

Sadly (because I took the opportunity of being in the US to make a couple of research related visits to companies on the West Coast) I missed the session on Tolerancing, that several people told me was one of the best discussions of the conference. But good to see this topic taking center stage.

So overall impression ….where is it going? In a nutshell less and less B-rep ….more and more GPU, meshes and point clouds…..and so what is needed is a representation that robustly spans all these domains. Not simply a flaky translation between them, but the geometric equivalent of the physics’ Grand Unified Theory (GUT) which can unite the worlds of scanned point-clouds, CGI meshes and the NURBS surfaces found in B-reps.  But where is the academic work that will lead to this Holy Grail?

Well of course the value of these meeting is not just the papers but also the chat in between the sessions and at one break I fell into a conversation about the “big picture” with a man who said “check out the paper on unifying sub-division surfaces with NURBS presented at this year’s SIGGRAPH conference (  http://doi.acm.org/10.1145/1531326.1531352 ), it’s a geometric GUT in waiting!”

Could sub-division surfaces be developed in a GUT that provides a common representation for all 3D modelling? Time will tell, but when you see the results for GPU based implementations of sub-divisions surfaces, such as that describe by Dr Baining Guo in his conference keynote ( see http://www.youtube.com/watch?v=3vrlJFOCbsM ), it appears to be a very plausible view of the future.