So last week I attended the 16th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (ACM GIS 2008), which was held here in Irvine. It's a little problematic attending a conference in one's own town — too easy to schedule other conflicting activities — but I still went to much of it, as well as getting some research done with attendee and renaissance man extraordinaire Matt Dickerson (about which more some other time).

One innovation (at least to me) was that the conference registration swag included a USB stick with the entire conference proceedings on it (and a key to use the local wireless network). I saw several people at the talks following along in the corresponding papers on their laptops. I don't know why the conference organizers also provided printed copies of the proceedings (I never broke the shrinkwrap on mine) but they did, and I heard at least one public call for them to continue doing so.

The research presented at the conference was an interesting blend of computational geometry, database theory (e.g. how to define and use consistency constraints for geographic data that may have different inclusion relations when described at different scales), sensor network data integration, transportation planning, geometric modeling, and other more applied problems such as pollutant plume reconstruction. I think the award for most creative presentation should go to Maarten Löffler, who also gave my favorite talk at Graph Drawing (a proof that any point set has at most one orientation for which it is the vertex set of a connected graph with axis-parallel edges, with Elena Mumford). This time, Maarten spoke about inferring the placement of unknown roads connecting points of interest to a known road network, by minimizing the dilation of the resulting completed road network. His paper has a huge number of co-authors, but all of them had some excuse for making Maarten go to the conference and give the talk. Maarten's talk slides used a roll effect to transition from one slide to another, making it appear that he was scrolling through a large map, the features of which were used as examples within his talk. Joachim Gudmundsson (who reported on the conference here) was also in attendance, making a whirlwind tour of the world that also included a Dagstuhl workshop involving some of Maarten's co-authors. He gave an interesting talk (though I missed the first part) on detecting common motifs in the movement patterns of moving objects, a problem that has applications in sports science, military espionage, and biology. In particular, he was looking to find groups of objects moving in single file, and had an approach based on Frechet distance for testing how well objects fit that pattern. I can't find an online version of the paper right now, but the title is “Detecting Single File Movement,” and it's with Maike and Kevin Buchin of Utrecht.

My own paper was with Mike Goodrich, and Mike presented it (though he gave me a bit of a scare by not showing up until five minutes before the presentation). The title is “Studying (Non-Planar) Road Networks Through an Algorithmic Lens.” Much past work on road networks assumes that they form planar graphs, but that's just not true; we present a different model based on sparsely-overlapping systems of disks, and show that this model still preserves many of the algorithmic properties of planar graphs. Mike also had a poster with Matt Dickerson on graph-theoretic two-site Voronoi diagrams, which again does not seem to be online yet. Afterwords, Mike asked me if I would attend future instances of ACM GIS at less-convenient locations. Most of my research is too abstract and theoretical to fit well at this conference, I think, but I enjoyed going to the conference and would definitely be interested in it as a place to send papers that are less about how the algorithms work and more about how to understand geography, as was true of our road networks paper this time.