- path planning
- scripting at the behavior-group level
- non-deterministic FSM's (probabilistic)
- constrained optimization
- social modeling (i.e. emotions and personal preferences)
What kinds of behaviors will we not be able to realize without these features? Should we include any of these?
I think when last we spoke, I suggested that evaluating such questions depended on the specifics of the scenarios you wanted to be able to simulate. In one scenario, social modeling might be key, in another not so much.
ReplyDeleteSo you could rephrase: "is social modeling necessary to solve the problem I'm tackling?" (is the problem you're addressing: all behaviors?)
Simplest place to start this big question: give me a list of a few scenarios you're hoping to address: e.g. pack chases prey, pair on collision course with pair in hallway, museum browse, checkout queuing...
The list (in progress) is here:
ReplyDeletehttps://wave.google.com/wave/waveref/googlewave.com/w+VBFR75zAA
I'm thinking about evaluation...
ReplyDeleteWhen you evolve dog pack behavior, can you identify traits in the existing results that you want your system to be able to match, or even improve upon? Are there any ways you might measure this match or improvement?
As you build the list (in progress) might a generalized description of the types of behaviors you are and aren't considering evolve? (e.g. "variants of animal following and chasing behaviors / not human reasoning-based behaviors)
Yes! We need to think about evaluation. We need to be able to write about our system in one (or all) of these ways:
ReplyDelete- our system provides improved visual quality or realism in some perceivable aspect of the crowd animation (subjective evaluation - hard to measure)
- our system is X% faster than this other system (objective and easy to measure, but not likely to happen)
- our system has a new capability or combination of capabilities that hasn't yet been realized either with respect to directability or generality of the crowd model (objective, assuming the background research is thorough enough)
Sure, I'll put a rough, generalized description at the top of the list and edit it as I refine the list of examples.