Line of Sight Lighting

The original FOW currently works by making non-explored regions darker, not by making explored regions brighter. So it's not preventing the dark side being dark, it's just that the original brightness contrast between day/night isn't very pronounced (yet).


As for the grayscale, I think it won't be too costly. I believe I just need to wait for them to allow some post-processing routines, and hopefully one will have access to the distance field information.

sounds good, i've liked your experiments so far.

Haha, thanks. It'll get a bit more interesting once we have access to more parameters in the shaders.

Greyscale requires the fog of war shader gets passed the current frame buffer (what would be on the screen just before the fog of war shader runs). B&W is how TA handled fog of war (apart from the dithered option), so it's been on my mind for doing the fog of war in PA.

I'm curious what you guys at uber will cook up for fog of wars. I've to say I always liked greyscale in TA, unless I played on a metal map. Grey on grey and my teamcolour was grey too. :mrgreen:

it sort of ruins the prettiness of the planet. if it was togglable that would be better.

Yeah but it's worse having the prettiness ruin the gameplay, fog of war needs to be fairly quickly identifiable, making the fog of war have less visual impact so the planet looks prettier is a bad idea IMO.

Well, it's a trade-off. I'd trade in a small amount of usability if it ends up looking fantastic.

The fog of war systems needs to be rethought all together. It's not only hard to visualize it in an aesthetic appealing way, but also has several negative effects on gameplay. It had them in TA, it had them in SupCom and it will still have them in PA.

Using hard borders is not only counterintuitive (which is why you even NEED a visualization in the first place), but also puts a harsh limit on the flexibility of the intelligence system. There are good reasons, why ANY other modern game, except for a few other TA inspired RTS and some "old school" RPGs, did away with fixed LoS circles for good.

I don't say that fixed range LoS wouldn't be nice to have as an fallback option if someone wanted to do a full TA conversion mod, but the engine should under no circumstances be limited to it.

My two cents:

I liked one of the grids from a page ago because it had a clear outline of all units' los, although the grid still looked kinda bad. Then I thought, is there any way to just take out the grid from beyond the outline, leaving a clear los outline?

Don't have alpha, so I don't know how this would look in the actual game, but I think a slim line would look clean while communicating its purpose very well.

I'm hoping these things aren't mutually exclusive, the day-night aesthetic is definitely a big selling point. I do like the idea of FOW being the atmosphere diffuse though. admittedly it might not give the same potential capabilities as grayscale. i haven't seen many RTS that use the grayscale for recency trick, i wonder why that is

Quite a few do, or at least a mix of darkened + greyscale with smooth transition between visible and non-visible areas.

Although none of them use the antiquated fixed range LoS system anymore. (Seriously guys, that thing is outdated!)

Most of them use real LoS tests so it actually looks more like some kind of natural shadows rather than an artificial border. It's just not that obvious and more intuitive, thats why you won't notice.

I think one of the reasons it has such a profound negative effect in PA is because it's not a flat surface, it's a sphere.

Lighting a sphere means you do not have a uniformly lit surface and that makes shadows etc look strange because you can see that the 'shadows' are not being generated by anything.

On a 2D surface this doesn't really apply because the entire surface is lit the same. You don't think about it because you don't see any areas with different lighting.

Agreed, and they aren't really mutually exclusive as long as the visuals don't affect the gameplay. There's a fair bit of room for improving the overall with look the contrast between night and day and I'm all for that, and Uber probably already plan to do something like that. But the second the visibility difference between night and day becomes noticeable (harder to differentiate between units) it begins impacting gameplay and IMO that's too far. There seems to be quite a few people who would like to push it towards that sort of level of difference because it does look much more awesome but I'm against changes to that degree, IMO the visuals can be improved sufficiently without going so far as to affect the gameplay.

Besides a bit of disappointment in the way line of sight looks, I'm a bit worried that when the game gets to the multi-planetary stage just scouting everything is going to create a huge amount of micromanagement... I think it is better to just let everyone see most things and hope they suffer a little bit of information overload and distraction than it is to hide things just because they don't have a unit standing close to it... As an example, when a enemy commander explodes I want to see it explode and the devastation it causes around it (to hopefully its own base), not a flash and a shadow as the unit that was attacking it gets taken out too.

There is one place I would like to be totally hidden though, the far side of any planet or moon where your army has no presence. The ability to build secret moon bases should be a fundamental right.

I second that.

The old TA system does not scale. It didn't in SupCom and it won't scale in PA either. Seriously, vision or intelligence all together in SupCom was a joke. You only ever needed it to uncover units for the radar and then you relied on strategic icons mostly.

Unless you had to micro something. In that case vision was quite helpful. Not because you wouldn't had perfect information anyway thanks to radar, but because it feels better to actually see the enemy, not just radar blibs.

Part of the problem with doing it as true LOS is most games can abstract their sight checks to a 2d height map shadowing. That can be calculated fairly quickly even with many units. For the unit counts we're looking to do it may be untenable to do actual line of sight just because of the performance implications.

I believe it's on the list of things to experiment with still though. The current version is literally just the first pass "get the feature in the game."

Congratulations.
This was the very first answer of any Uber employee on the concerns about the usability and scalability of the TA/SC style vision/radar for over 6 months.

But it doesn't need to be real LoS anyway. Not only because LoS test is expensive and it would require to test each unit against each other, but because it isn't necessarily required, although nice to have as an option for small games.

Even some type of graded intelligence works better than the TA system, and even if it was just some primitive formula like visibility = ((unit size - range) > vision). Given a not completely-messed up data structure, this is still possible to calculate in O(n²) worst case but O(n log(n)) on average. There have been many suggestions on various possible systems which all work with reasonable computational overhead and provide various appealing options for intelligence, some even managed to solve the issue that intelligence wouldn't scale with the ressources thrown into it.

Sure, none the is as fast as an octree based two pass implementation of purely static vision radius which will do O(n log(n)) even in worst case and can even go down to o(n) in best case, but thats something which needs to be tried.

That one had me chocking in laughter. You got a Monkey Island reference in PA.

You're already relying on cubemaps for pathfinding. Perhaps a cubemapped heightmap kept in sync with the CSG would be adequate optimization? Lots of corner cases, though.