The basic idea of a motion graph is simple, but there are about a zillion variants.
The required reading is either our 2002 motion graphs paper or (recommended) the chapter on motion graphs in Lucas’ thesis, a skim through my bibliography of pre-2008 papers, and at least one additional paper. The thesis has the nice advantage that it was written a year or two later, so we could compare with other methods better and improve some of the details. There is a Moodle question to answer.
- Automated Methods for Data-Driven Synthesis of Realistic and Controllable Human Motion. Lucas Kovar. PhD thesis from Department of Computer Sciences, University of Wisconsin – Madison – 2004
The idea of concatenative synthesis (of which motion graphs are one of the simplest forms) has been around for a while. In 2002, there was an explosion of interest and several groups published papers where the graphs were built automatically. I think our paper had the easiest to understand explanation, and each of the papers were subtly different. Over time, the differences between the methods proved to be less important than their similarities (so, it’s not really worth spending time on them). If you read Lucas’ thesis, he does a bit of a comparison.
- Motion Graphs. Lucas Kovar, Michael Gleicher, Frédéric Pighin. ACM Transcations on Graphics, Volume 21, Number 3, page 473–482 – jul 2002.
- Interactive Contol of Avatars Animated With Human Motion Data by Jehee Lee et al. SIGGRAPH 2002
- Interactive Motion Generation From Examples Okan Arkan and David Forsyth, SIGGRAPH 2002
After these, there was an explosion of motion graph papers. And it turns out, there were earlier papers we weren’t aware of. I made an annotated bibliography of many of the early papers. MotionGraphBib2008
You should pick at least one paper from there to read. However, I strongly recommend:
- Snap-Together Motion: Assembling Run-Time Animation. Michael Gleicher, Hyun Joon Shin, Lucas Kovar, Andrew Jepsen. ACM Transcations on Graphics, Volume 22, Number 3, page 702–702 – jul 2003
Because it highlights some of the issues in the early motion graphs work, and has simple enough techniques that you could try to do them for a project. And, these methods are a lot more practical than some of the fancier newer stuff.
We’ll look at the fancier newer stuff later in the class.
