RoadsAfar

Dryad is a new tool from the Stanford Virtual Worlds group for creating trees intuitively. The demo is very polished (particularly for something worked on by graduate students) and the idea is promising in its application to other domains. Vladlen Koltun gives an excellent demonstration of the technology in a presentation.The technology works according to the following steps:

1. Create an algorithm to generate a tree according to a set of predefined parameters.
2. Build an editor that allows the parameters to be modified by some intuitive controls (sliders, drop-down menus, etc.)
3. Turn over the editor to users and track which trees the user saves or exports (this counts as a “vote” for the “goodness” of the tree).
4. Generate new random trees, favoring random trees with a lower distance to “voted” trees in the high-dimensional space.

There are several promising aspects of the technology that are worth mentioning in particular:

• User-defined design space exploration- As Vladlen mentions in the video, the vast majority of trees in the design space are simply not tree-like. As users played with the program and created their own trees, the trees that are saved are used as votes for the trees goodness. This is reported to be similar to the way that voting will happen using the editors in Spore, Will Wright’s new game that is built around several procedural editors. It’s a powerful idea for determining the pockets of goodness in a sea of procedural randomness.
• Map analogy for space exploration- After the space has been parameterized and the areas of the design space containing “good” trees has been determined, the user is allowed to explore the space on a two-dimensional map-like landscape similar to Google Maps. This analogy is familiar to users. Furthermore, as the user zooms in on an area, they are exploring more closely a particular region of the multi-dimensional design space and the trees in that space resemble each other more and more closely the further the user zooms in. Others should take note of this method of design space navigation.

Kudos to the Stanford group on the tool and the polish behind it. I look forward to seeing what objects are explored next. 

“…find (a) simple solutions (b) to overlooked problems (c) that actually need to be solved, and (d) deliver them as informally as possible, (e) starting with a very crude version 1, then (f) iterating rapidly.” -PG