"I, Sir, have had several speculations in my pants... I'm not going to rant about them again... I need to grip more to explain my ideas
(blip: Tensor Automatons -- not exactly fields -- is the name of one of them, probably the most relevant one)
Well, no need to rant, but it's hard to argue against something that one cannot see
When you've formulated your ideas, we can discuss them concretely I suppose?
Meh. Dimensionality can be good sometimes. I guess you're right in this case, though. I really favor a single collisions implementation over specializations. OH YOU, ADDING COMPLEXITY LIKE AN EVIL SIR. (teasing you)
Oh no I completely agree with you, I avoid specialization at all costs
Just saying that, realistically, heightmap collision can be substantially accelerated by taking advantage of its structure. But personally I probably wouldn't specialize it either, for the sake of generality.
Oh, is your collisions engine voxelizing meshs and using 'le voxels? I guess that'd be unified collisions with a scalar field.
Right now it's using a triangle-based algorithm, but I will probably change it later to directly use the fields so that it can run on the GPU. Mathematically, unified collisions under a field is absurdly simple: two functions, f and g, find x such that f(x) = g(x) = 0. Nice!
False. False false false.
Well you're going to need to elaborate a bit, because wikipedia, the ever-present argumentative force, is one my side...(
http://en.wikipedia.org/wiki/Electric_potential http://en.wikipedia.org/wiki/Gravitational_potential). Sorry, I'm not trying to insult your intelligence by linking articles to potential..but without explaining your objection...it's the best I can do
Me wrote:I'm interested in establishing a framework to dynamically reduce explicit/classical representations into various orders of abstraction with intent to simplify computation necessary for the simulation of systems. If you prefer that I specify how I imagine these systems might be, consider a very large set of points which may exhert various forces upon each other (classically, like particles) within the system e.g. exhibits contact dynamics and supports modelling external forces and fields from outside the system of points. Consider a large number of these points behave like a fluid, and they're contained in a way that can be approximated using a planar-surface fluid approximation method. I haven't researched these enough yet (planar fluid approximations), but I think it's an interesting idea to explore means to adaptively approximate systems using poweful mathematical generalizations. Of course, this particular type of fluid approximation I just gave for example is very limited (planar), but that's why I'm interested in adaptiveness; specifically a broad framework of malleate abstraction devices that can effectively perform these adaptations.
I like to think of this strategy like the theme of the book "A Wrinkle In Time"; virtually traversing spacetime by welding two points together into one. Rather than brute forcing the simulation, congruences may be identified (transitive actions, linear dependence, fractal phenomena etc.) to yield shortcuts and inherently simplify computation. You may call this an adaptive model or approximation.
I'm thinking of some kind of "tensor automaton, " if you can imagine what I mean by that (think of an intelligent lego sculpture). Are you aware of any ideas or do you know of information that is relevant and maybe interesting to this topic? Please discuss your thoughts.
It sounds great, but I'm not sure how it is relevant to the applications to which I've spoken of applying fields. You're speaking of simulating systems, I'm speaking of using fields as mechanisms to define and control a behavior (i.e. a surface, a collision interaction, AI movement, etc). I'm not seeing the conflict of ideas? There is no large-system simulation involved in any of the applications I've mentioned. Really, I'm just talking about the
underlying representation of the algorithms used in the game, and how a unified representation can be exploited to certain ends. But if you have a better representation, I'm all for it, but it sounds as though you're speaking of developing accelerators for computing interactions, which is not really what I'm doing.