for locomotion there seems to be 2 strategies - cell-based or continuous space - similar to how
fluid dynamics is modeled: Euler or Langrangian. in fluid dynamics there is also a hybrid approach that might also be applicable to crowd simulation.
so by locomotion, you are planning on dealing with collision avoidance, collision detection, and collision reaction. there are also issues with 2-way traffic in corridors, passing someone going the same way, how to keep groups together under limits to navigation (e.g. in a corridor).
is this the level that you plan to identify issues?
Yes, that is what I was thinking. I'm hoping to solve the navigation problems of locomotion, but I'm also hoping the tool will give the designer control over some expressive qualities of the movement.
ReplyDeleteI was planning on taking the Lagrangian approach, do you forsee any problems there?
- J
of course neither approach solves all the problems. the problem with L is that 1) crowds become an n-body problem and 2) interaction with environemnt has to be computed. E solves these by essentially a bucket sort but at the expense of restricting the field of interest to the buckets. Of course with L, you can use various sorts to help the n-body comparisons and pre-process the environment to help with interaction.
ReplyDeletein locomotion, there is also the issue of how to handle 'vision' of the agents: omniscience? or are you restricting knowledge of the environment at the agent level?
Yes, we'd need to use some sort of adaptive spatial sorting to prevent n-body comparisons, and I had intended to use the Reynolds method of vision (an area of vision around the individual defined by a radius and an angle) but this is totally up for experimentation.
ReplyDeleteI'll need to dig into some more of the crowds literature to get a feel for whether E or L seem to be the commonly used approach.
I think both are used. E is used when the space to be considered is very well defined and when interaction or large populations are needed - I've seen it, for example, in modeling panic situations in buildings (a common, seemingly well-funded application). L is used more for off-line calculation and/or the space is not known apiori
ReplyDeleteWell, we do need high levels of interaction and the space could be well known apriori (assuming the designer defines the space at the beginning of the process, i.e. the layout and scene modeling has already taken place). Our calculations are technically off-line though, even though we like to have fast feedback.
ReplyDeleteDo you think Euler would be a better approach?
well, a pure E approach, agents go cell-to-cell, is usually when only tracking location is important. if you want good visuals, then you probably need to be working in the continuous space of L. you might use an E-like bucket sort to reduce the n-body comparisons - plus you're talking of relatively small 'n', so I don't know if E would be that useful.
ReplyDeleteOkay. Sounds good. I'll keep heading down the Lagrangian route, and we'll see what kind of spatial sorting we need to do when we start tackling collisions.
ReplyDelete