is there a difference between a crowd that is, say, in a market square versus a crowd that is an invading army versus a crowd that is an audience (like in a stadium or movie theatre).
what are common among *all* crowds (multiple agents - is that it?)?
what are features, if any, that differentiate crowds in meaningful (for animation) ways?
transient-ness? uniformity in mission? density?
what about viewing distance? are individual actions important? are the actions only statistically important? do the actions even need to be distinguishable?
what's the difference between a crowd and a large number of 'intelligent' agents? anything?
is your objective to be able to model all crowds in all situations - or are you identifying a subset of crowd type to focus on?
I'm not sure that all crowds can be modeled using the same framework, but I'm hoping for as much generality as is reasonably possible. One limitation, that I've intentionally engineered into the proposed framework is that my model only supports crowd locomotion. I haven't made allowance for (and would rather not implement) motion clip blending and all the problems/infrastructure that come with it. It seems to me like that is a heavily studied area, and that I would not easily be able to add to the body of knowledge without significantly more time + effort.
ReplyDeleteSo I've intentionally made my crowd framework only deal with the locomotion of the crowd individuals and not their specific animation cyles. That means we can't model the "audience" type of crowd, seated in a theater. But hopefully we will be able to tackle any type of non-stationary crowd. Do you see any obvious limitations of the proposed framework in that respect?
It seems like the matters of density, viewing distance, and the importance of individual actions come down to the scale of the crowd in question. I was hoping to aim for a tool that choreographs a mid-sized crowd (less than 50 but more than can be easily keyframed, say 8). That would mean that the actions of individuals must be coherent and that the viewing distance could be relatively small. The density depends on the physical environment occupied by the characters.
I think a crowd for this purpose is a collection of "intelligent" agents. Do y'all agree?
As far as features that differentiate between different crowds, I think we should come up with some sort of list so that we can use it to (subjectively) evaluate our phenotypes so we can tell if we are getting the range of variety out of our model that we want.
- density vs. sparsity
- uniformity vs. scatteredness
- cooperating vs. antagonisticness (I'm making up words here...)
- homogeneity vs. heterogeneity (w/ respect to any number of attributes: speed, turn rate, etc...)
- relating to one another vs. aloofness
- and so on...
Yay! comments with notifications!
ReplyDelete(Just wanted to let you know it was working...)
In trying to evaluate your thoughts above, I keep returning to wanting to know the specifics of the problem you're hoping to solve.
ReplyDeletee.g. If you say you want crowds of roughly 8-50 people, then battles, festivals, mobs, audiences, rallies, etc are all NOT part of what you're addressing. So this makes me think about what you are addressing... cocktail parties, meetings, street scenes, architecture traversal, games... what else?
Hmmm. Has to be things where the primary visual evaluation is only position and orientation you said, right? Not posture based.
Could you perhaps put together a list of examples of scenes that a 3D animator might need to craft crowd locomotion paths for... and then we could use the list to inspire the range of "controls" one might one for things like density vs sparsity, etc?
While a list of scenario descriptions would be a good starting place, even better eventually would be a list of youtube URLs showing example clips of such crowds and qualities.
Great idea! I'll put that on my list of things to do.
ReplyDelete