Automatic navigation meshes

[aegis]

it's shown to work with a large number of units running around (army retreating over mountains, then a larger army following), or fast-moving large units (remember the UFO zooming around the city in the demo video?). attacking someone else's point doesn't make yours any more right

[Argh]

The simulation you were shown doesn't actually demonstrate either case, so perhaps you should consider what I've said a bit more seriously and do that napkin math. The answer is kind've important.

[aegis]

how do the provided examples not match the efficiency in question?

and if you're absolutely certain the aforementioned napkin math will prove your point, don't tell me to do it, post your (specific) results!

[Argh]

Fine. Napkin math says that:

1. Size increases practical workload by n^2. You can't project a single line and arrive at adequate results using heatmaps, otherwise you have large agents constantly bumping into things they can't "see".

2. Velocity increases practical workload for checks for best results somewhat larger than linear (due to a conic optimization for projective analysis, if you want it to be smart enough to seek low-congestion zones early) and the trig workload also follows this trend.

In short, besides the other practical problems with heatmaps (topological inexactness, high static data load, relatively large numbers of sectors to reevaluate per cycle) there are serious complications when the actors start to exit the normal case, as is typical in a RTS like Spring, where we want wildly different sizes and velocities to be supported.

[aegis]

1. so anything with a size at all has a greater workload of n^2 than things that do not? your statement is a conclusion without any relative data, which invalidates your initially prospectivated hypothesis.

2. can't argue with number two either!


3. tell me how the video does not demonstrate both size and velocity with the ufo, or scale and optimizationses in a general sense with the armies. also, download supcom 2 (or be lazy and lookup youtube videos of their pathing) as an example of this system working in real-world rts situations

[Argh]

1. so anything with a size at all has a greater workload of n^2 than things that do not? your statement is a conclusion without any relative data, which invalidates your initially prospectivated hypothesis.

That's because I don't need to provide any data, it's obvious. A four-square agent needs to check more squares, influences more squares, and so forth.

You can't argue with 2 because you have no idea how what they're doing works, apparently. They're just checking ahead to see the heat values per zone. You need to do some trig to determine which zones to check, and use a triangle to get the best choice when velocity gets high enough, because you want to make choices earlier to avoid having agents braking when it's not necessary. Again, this should be obvious.

As for the UFO, note how far the forward influence had to be, and watch carefully as to size. Moreover, that's a special case, where they wanted it to push the simulated agents around and make them respond to the UFO. The UFO is clearly tolerant of congestion. The system I'm thinking about would not preclude that sort of behavior, but for various reasons it's probably not something we want in a RTS. Instead, in a RTS we want congestion-tolerant agents that use local steering to deal with congested zones, to save some steps and allow for stuff like a Krogoth surrounded by a sea of Peewees.

As for the armies simulation, the behavior of what I'm thinking about would be nearly identical, but with a lower cost.

[aegis]

supreme commander's "new pathing system" demo specifically demonstrated three large units walking one way and a formation of little fast units walking the other

[Argh]

Nothing I've said here says that what I'm thinking about won't allow for that. I can't go into why without giving you information that would not be appropriate until I can make a full presentation.

[aegis]

Nothing I've said here says that what I'm thinking about won't allow for that

but you are saying this pathfinder would have problems with it, when it demonstrably does not.

create a working proof of concept and directly compare it to the pathfinder in supreme commander two. this will be more appropriate than refusing to give details, or providing napkin math which is actually unbacked rephrased versions of your previous points.

[Argh]

but you are saying this pathfinder would have problems with it, when it demonstrably does not.

No, I'm saying that it has to spend a lot more CPU than we ideally want to arrive at that solution.

[AF]

The benefits of crowd continuum are precisely improved performance in both real terms and simulatory outcome.

The problem is your arguing that your theory will be faster and better and that crowd continuum is slow, despite no evidence supporting your claims, and a body of research, practical demonstrations, and commercial applications weighing against you.

If you Aegis were to post on the forums that he has devised a new theory of subatomic physics, and quantum mechanics is a load of nonsense, you would be naturally suspicious, and want proof, you cannot expect us to do the same to your theory when we have yet to even see a basic summary of it nevermind a demonstration.

So far all we know are the benefits of your algorithm yet we have no details at all on its working or the principles behind it. What's more since it has not been implemented, you do not know if it will indeed pan out the way you intend. I've implemented algorithms before and found them to be slower than I expected, and that certain optimizations help a lot despite appearing to be slow and clunky in my head, you have to trust that your world model internally is not fully representative of the external world and corrections need to be made throughout your life.

For those who have yet to see, the video can be found here, along with the paper on crowd continuum in PDF form of 9 pages:

http://grail.cs.washington.edu/projects/crowd-flows/

[Argh]

So, basically talking further about this here is a waste of my time. Tired of this.

[aegis]

no, purposely talking *vaguely* about it is a waste of your time

[Argh]

I explained why I do not want to give anything other than a full description.

You guys have persisted in being jackasses about this for two days now. If you think that makes me want to spend an hour or two doing a writeup and illustrations and explaining the algorithmic approach I have in my head, you're wrong.

Next time, when I say I don't want to talk about something yet, just drop it.

[zwzsg]

Oh by the way guys, last time at lunch I had five minutes of free time so I mentally solved Riemann conjecture. Sorry, I don't have the time to write down the demonstation, I'm busy tackling another issue. I can't tell you what that other issue is, but it involves death by incontrolled cell multiplication. Let's just say it should have fairly big implication in term of healthcare. Anyway, my cure should be ready in a day or two, I'll say more then.

[aegis]

If you think that makes me want to spend an hour or two doing a writeup and illustrations and explaining the algorithmic approach I have in my head, you're wrong.

Next time, when I say I don't want to talk about something yet, just drop it.

1. you spent enough time talking about it, you might've saved yourself a few posts and showed us something cool in one swoop.

2. you talked about it an awful lot for not wanting to talk about it yet

[AF]

Your meant to post the gist of what it is your thinking so that we start saying things like "wow that sounds interesting tell us more"

Instead you did what you did and you activated a set of routines in every readers mind that we have been trained by the internet to do:

"Filter out flash banners, be suspicious of sales talk, and delete spam emails. I dont have time for the usual omglookatmyamazingproduct speak, I got a Wow subscription to justify and 30 wikipedia tabs I havent read yet"

People here like substance, if you promise substance soon then fine, if you promise substance then talk about how great this substance is, but never actually provide it, dont be surprised when scepticism sets in.

[zwzsg]

People here like substance, if you promise substance soon then fine, if you promise substance then talk about how great this substance is, but never actually provide it, dont be surprised when scepticism sets in.

Fair enough, I will now tell you what my substance is, and how it works:

1. The treatment provides complete elimination of any malignant cells.

2. The treatment provides equally for primary tumors and metastasis, as well as leukemia.

3. The treatment works faster and with less side-effect, requires fewer and less invasive interventions than current methods, and will be very efficient in every case, thanks to various improvments.

4. The treatment provides cheap solutions for massive populations, and other common problems that defeat traditional approaches.

5. The treatment can be used in a variety of illness, but is optimised for carcinoms.

Now that I gave you all the nitty-gritty technical details, can you please stop harassing me?

[Guessmyname]

...

I am regretting starting this thread.

[Argh]

Oh by the way guys, last time at lunch I had five minutes of free time so I mentally solved Riemann conjecture. Sorry, I don't have the time to write down the demonstation, I'm busy tackling another issue. I can't tell you what that other issue is, but it involves death by incontrolled cell multiplication. Let's just say it should have fairly big implication in term of healthcare. Anyway, my cure should be ready in a day or two, I'll say more then.

Something like that, but with less sarcasm.

It really annoys me that you guys want a haiku version of something that can only be presented in full form, if I want it to be properly attributed in the future. So bugger off, please.

I've said why I can't do that right now. I've given you enough hints that if you were interested, you could even pursue the same routes. It's just common courtesy to assume that I will get back to it when I am ready to provide full explanation.

You're more than welcome come up with innovative solutions, if you are able to do so. But pestering me to provide you with an incomplete explanation is totally futile, and just pisses me off.