So I finally opened Geppetto today and felt the need to say something.
Cool idea/start.
First, an observation/question: I noticed that the XRefed models are using Biped. But I've been under the assumption that AD is moving towards the use and development of CAT. Will this always be tied to Biped? Will it be loose, allowing Biped/CAT/Pinochio or whatever anyone else wants to use?
Now to ideas.
I'm assuming that the controls behind the scenes have functions/weights that create behaviors that must be granular at the calculation level... meaning that Person 1, although following a general "flow" must choose to slow down/ go faster to avoid collisions. It would be really cool if the user has access to those rules... such that you can pick Person 1 and animate those controls over time--and make it such that manually animated behaviors supercede the generic flow and forces the other automated/non-animated people to adapt/adjust.
Similarly, it would be cool to be able to assign the people to Groups that also have controls like that--so you could animate the crowd. Let's say that a car collides with another car at one end of the simulation... the crowd nearest needs a set of reactions that is different than other crowds further away. Alternatively, expand the Idle Areas to generic Groups that are then holders for the parameters of that group--instead of just Idle... this area could be controlling whatever arbitrary parameters the crowds will eventually have--and preferably be animatable (people slow down automatically in this area starting at frame 212 because that's when the puppet in the window starts to grow a really long nose! and then walk away when the curtains close at frame 2400...)
Finding ways to integrate the behaviors with CATMotion would be really cool. If you could create a collection of CATMotion (and other layers) to be applied to the behaviors that crowds could have... that would be awesome. Or perhaps find a way to make the XRefed file contain the controls... so that XRef A (biped) exposes behaviors 1,2 and 3 for its model in the scene and XRef B (Cat) exposes behaviors an entirely different set applicable to its participation in the flow. Exactly how to do that in a unified way... I'm not so sure... but it's an idea.
Anyway, these are my initial thoughts.