Tree branches contact - Drag or damage in Battle of Britain?

[BG-09]

Hello all!
In Il2 series, I remember any contact with the forest trees and their branches was leading to crash.
I have seen Vietnam war documentary, where, the rotor of the Bell helicopters in low level jungle flight, cut entirely tree top branches thick as 15 centimeters - that is confirmed by pilots. So similar is tree contact with the aircrafts.
I have read memories of Ulrich Rudell, who said, that his Ju-87, diving with heavy bomb load in the dark conditions, have cut entirely with the both wings - left ant right wing, two tree tops 10 centimeter thick, and the tree tops remain hanging below the wings of the aircraft. As You see, there was no instant crash. Ulrich Rudell land his aircraft at his airfield successfuly.

The same famous pilot, Ulrich Rudell, afther short take off from one castle yard, some kind of lawn flying with Fi-156 collided with giant oak, and the poor Fi-156 remained on the top of the tree intact. Grownd crew successfuly helped to Ulrich Rudell to come down from the tree.

What do you thing?
I belive, scalable damage for the aircraft must be modeled in case of tree contact.

And one more thing - trail of broken trees if You fall in to the forest....and why not forest fire.

Oleg, please do something...

regards to all pilots!

BG-09

[Tbag]

I'm sure that not every contact with a tree ends fatal. But what you propose would mean that each tree needs a kind of damage model. And the question is if that effect is worth all the programming effort and if it is worth the cpu rescources. I think there are far more important things but than again I'm not a lumberjack :D

[BG-09]

Tbag, I mean damage model for the aircraft, but not for the trees. : )
I repeat: We do not need any damage model for the trees.

regards!

BG-09

[bomath]

To have that kind of contact modelling the tree would need DM too.
I'd be happy to actually *see* the forest when I'm flying @ 10-30m altitude; right now I can see individual trees (if they're lonely in the middle of the terrain) but the forrest is depicted as horizontal textures, even with forest=3 in conf.ini; the horizontal textures obviously have no width/depth, so when you also fly horizontal and very close to the ground... BOOM!

[JVM]

I do not think you would need a damage model for a tree: I believe it would just be necessary to know the height of the tree, the height of the A/C and you just have to cut your tree in say three parts:
- height of A/C in the upper part of the tree = big noise, everything OK (believe me I experienced that first hand with a PA-25!); if two trees or more, the flight stops right there, not necessarily with a boom...
- height of A/C in middle part of tree (this would be actually a very small part, and maybe not even existing) = big hole equivalent to 20 mm explosion, consequences on handling, still flyable with skill...if two trees or more, boom
- height of A/C in lower part of tree: boom

Add in the picture coefficients to take in account the nature of trees (pine not equal to oak for that matter!) and you could be done...

Even a simple separation in two zones on the tree, without the complexities as described above would add a lot to immersion!

JV

[skarden]

I do not think you would need a damage model for a tree: I believe it would just be necessary to know the height of the tree, the height of the A/C and you just have to cut your tree in say three parts:
- height of A/C in the upper part of the tree = big noise, everything OK (believe me I experienced that first hand with a PA-25!); if two trees or more, the flight stops right there, not necessarily with a boom...
- height of A/C in middle part of tree (this would be actually a very small part, and maybe not even existing) = big hole equivalent to 20 mm explosion, consequences on handling, still flyable with skill...if two trees or more, boom
- height of A/C in lower part of tree: boom

Add in the picture coefficients to take in account the nature of trees (pine not equal to oak for that matter!) and you could be done...

Even a simple separation in two zones on the tree, without the complexities as described above would add a lot to immersion!

JV


That i think,would have to be the best solution that anybody could really come up with.

+1

[BG-09]

I do not think you would need a damage model for a tree: I believe it would just be necessary to know the height of the tree, the height of the A/C and you just have to cut your tree in say three parts:
- height of A/C in the upper part of the tree = big noise, everything OK (believe me I experienced that first hand with a PA-25!); if two trees or more, the flight stops right there, not necessarily with a boom...
- height of A/C in middle part of tree (this would be actually a very small part, and maybe not even existing) = big hole equivalent to 20 mm explosion, consequences on handling, still flyable with skill...if two trees or more, boom
- height of A/C in lower part of tree: boom

Add in the picture coefficients to take in account the nature of trees (pine not equal to oak for that matter!) and you could be done...

Even a simple separation in two zones on the tree, without the complexities as described above would add a lot to immersion!

JV



Realy good proposition JVM - sounds very reasonable. easy and relistic decision of the problem. You know, that colision with the part of other aircraft in Il2 series, leads to desintagration of your aircraft or both aircrafts - similar effect must appear be in the case of colision with tree branch.

JVM, say something more about your experience with PA-25 please!

Regards!

BG-09

[X32Wright]

It might look simple on paper but in the 3d world thats another issue. If this is implemented even with LOD (level of detail) geometry, this would lead to an increase in the geometry load since every breakway section would need to have a separate geometry (plus the LOD geometry) plus bounding boxes (for collision detection). Even with adaptive subdivision this multiplied by thousands of trees would be a nightmare. IF its possible at all it would restrict online play to maybe just 6-8 people nothing more due to the amount of data that needs to be exchanged as well as the amount of data that needs to be processed.

Even if they use flat geometry and simple geometry (triangles) ovelapped with Alpha-channelled textures this would still lead to more geometry and more texture maps to load during gameplay. This would be impossible. Using shaders (Cg or rendermonkey) might help but still it would be too much for online play.

[Codex]

You need to Broaden Your Horizon fellas ... get it ... Broaden Your Horiz....ah forget it. :razz:

Have a look at this: http://www.gamedev.net/community/forums/mod/journal/journal.asp?jn=263350

This is terrain generation procedural style. It is CPU intensive but way less expensive on storage space i.e. RAM, but with multi core CPU's pretty much main stream now who cares. So with the saving in RAM bounding boxes could be cached to aid in collision detection between tree and planes.

[BG-09]

You need to Broaden Your Horizon fellas ... get it ... Broaden Your Horiz....ah forget it. :razz:

Have a look at this: http://www.gamedev.net/community/forums/mod/journal/journal.asp?jn=263350

This is terrain generation procedural style. It is CPU intensive but way less expensive on storage space i.e. RAM, but with multi core CPU's pretty much main stream now who cares. So with the saving in RAM bounding boxes could be cached to aid in collision detection between tree and planes.

Thanx a lot Codex for the link! Extreeeemly good graphics...

I belive it is not dificult to model tree colision. New CPU's are coming...

[X32Wright]

You did notice that those trees in there were made with at least four (4) triangles each (if it wasn't subdivided into segments) with a texture map right? If you want some interactivity the number of polygons would need to be at least doubled if not quadrupled for each tree. Still a big problem if you want collision detection with planes and trees. The additional geometry load in the game is really much better being used for the planes and interactivity between planes in the game.

If you are so concerned about trees then I think you are flying way too low and getting damaged too often :)

[BG-09]

If you are so concerned about trees then I think you are flying way too low and getting damaged too often :)

As You may be know, famous proverb in our community is this:

"...at altitudes at which this community is flying, diving is not an option..."
So not only I but many fellow pilots do have contacts with the trees.

Oleg is perfectionist, so the problem will be terminated.

I am not programer anyway.

BG-09

[|ZUTI|]

I think that we don't need the trees breaking at all, just a bit better DM when contacting a tree. If we can live with what we have now, we would sure be able to live with what we could have then :) And if tree makes a plane crash, just replace it with a burned tree texture ;)

I just don't want the kind of forests that we have now (3 layer textures). But from what we've seen so far, forests will be forests. That's cool.

[X32Wright]

Sure it would be possible to have trees with DM but as I said multiplayer number would have to be reduced to 6-8 at the most nothing more. Forget about doing 30-50 players with each player having like 5K polygon load each even with LOD. Thats crazy!

[Codex]

The poly's of the trees would be done on the client PC's, it would have nothing to do with the server, all the server would have to worry about is the events on the map.

I just wonder if Oleg was to 'use' some of graphic tricks from the current FPS game engines (Unreal, Doom Crytek) that a lot more could be done in terms of graphics, what I suspect is that are certain patent issues to get around.

[proton45]

It seems to me that you would only need to place the high count polygon (damage modeled) trees near airports and landing strips....maybe just place them at the end of runways.

I don't think that every tree needs to be modeled in such detail, but it would be nice if you could place these "tree obstacles" in limited strategic places...

Maybe mission builders could place the "high-poly" trees around ground objects that need to be strafed in the course of a mission.

Other ground obstacles that you commonly find near airports like telephone poles could be placed in limited areas too...

[X32Wright]

The placement of the 'high resolution' polygons are not determined by where they are 'positioned' in the map but rather how close they are to the 'virtual camera'. Although this cane be done too by the game developer meaning it can program the higher density polygons to be loaded near the bases but in general LOD (level of detail) loading is determined by view and camera distance. So this means if you are looking 'externally or 'from inside the cockpit' this distance to the trees determines HOW and WHEN the high resolution polygons are loaded. The issue is still the same because even with LOD too much detailed trees would still increase the overall polygon load specially with damage modeling on them.

Surely 2D polygon cutout trees as modeled and rendered can be made to orient to the camera hereby making it appear 3D, it is the number of them thats still an issue in a fli-sim. This 'polyface orientation based on camera view' technique is already being used int eh game though in the cloud generation.

FPS games tho cheats, they dont exactly show all the trees or all the polygons since there is a lot of CLIPPING and CULLING (back-face) happening there based on occlusion. This means objects that are obscured even temporarily by the camera are immediately removed or buffered so in these fps games they always look like there is more in the scene than actually is being 'rendered' in real-time. This is harder to do with fli-sim where your view extends 'forever' at at least as far as the camera near and far clipping planes are set to.

[JVM]

If it was possible indeed to use the technique I described higher in the post, technique which involves only a computation based on relative heigth, it would be the cherry on the cake to make one (or two, varying the cut heigth) damaged tree model (= cut normal tree...would you believe that?) for each tree type and

- either to replace entirely every tree impacted in the crash by its closest damaged counter part...
- or to make it simple remove the impacted trees and then no need to make a damaged tree!

Both would be visually quite good...Now what impact would it have on the FPS, compared to a normal crash with associated crater for example?

This being said I hope the craters will be a lot better and more complex than in IL2...whether they are from bomb or crashes...the Il2 ones are ugly and nowhere near realistic, even for a texture! I have never understood why? Immersion killer again...

JVM

[X32Wright]

I dont think you understand the enormity and complexity of tree interaction and the resulting increase in polygon load. Surely there would be programmed to have localized adaptive subdivision when it comes to collision detection based on proximity of a plane to the ground and trees but those would still require enormous resources (CPU,GPU and polygon load). Most GPU today are not made for handling enormous polygon loads but rather are optimized for enormous TEXTURE handling instead as well as shader processing and some geometry calculations. They aren't geared for handling 'interactive' 4 million polygon scenes at all (although they might be able to load that in memory but not interactively) while the CAD workstation cards are (which renders everything accurately and NOT cheat like the game cards which do not render everything), although things are changing.

Computation based on the height will work since they trees and such are placed on 'height field' generated terain as it is already but even with a simple 4 triangle tree that would require more resources and more polygons to pull off IF you want interactivity with the plane instead of a simple collision detection that we have now. I am asuming that the collision detection (plane vs ground or trees collision detection) in the game isnt even collision detection per se but rather based on the coordinates of the plane's axis. IF the computed vector (for the plane) falls beyond the 'ZERO ground axis' the plane is designated as crashed. But this is a guess. Having to apply this to trees is a waste of time. This is why we crash on the large 'invisible trees' (which are 'height displaced geometry' (pancaked) mapped with alpha channels to look like forrests) but pass through the ones that look and feel solid (those with swaying detailed leaves).

As for better craters well, Oleg can always implement bump mapping and procedural displacement (even animated) but didnt chose to do so in Il-2. My guess would be due to the video cards' capability limitations when Il-2 was released.

[JVM]

I dont think you understand the enormity and complexity of tree interaction and the resulting increase in polygon load. Surely there would be programmed to have localized adaptive subdivision when it comes to collision detection based on proximity of a plane to the ground and trees but those would still require enormous resources (CPU,GPU and polygon load). Most GPU today are not made for handling enormous polygon loads but rather are optimized for enormous TEXTURE handling instead as well as shader processing and some geometry calculations. They aren't geared for handling 'interactive' 4 million polygon scenes at all (although they might be able to load that in memory but not interactively) while the CAD workstation cards are (which renders everything accurately and NOT cheat like the game cards which do not render everything), although things are changing.

Computation based on the height will work since they trees and such are placed on 'height field' generated terain as it is already but even with a simple 4 triangle tree that would require more resources and more polygons to pull off IF you want interactivity with the plane instead of a simple collision detection that we have now. I am asuming that the collision detection (plane vs ground or trees collision detection) in the game isnt even collision detection per se but rather based on the coordinates of the plane's axis. IF the computed vector (for the plane) falls beyond the 'ZERO ground axis' the plane is designated as crashed. But this is a guess. Having to apply this to trees is a waste of time. This is why we crash on the large 'invisible trees' (which are 'height displaced geometry' (pancaked) mapped with alpha channels to look like forrests) but pass through the ones that look and feel solid (those with swaying detailed leaves).

As for better craters well, Oleg can always implement bump mapping and procedural displacement (even animated) but didnt chose to do so in Il-2. My guess would be due to the video cards' capability limitations when Il-2 was released.

I may not have explained myself clearly: in the system I was thinking about, the collision detection is between a segment (representing he tree trunk at the tree geographical position) whose length and altitude is known and the parts of the A/C model used in the collision detection. I agree with you in that no additional part of the tree and notably not the triangles giving it volume (assuming the trees would be real individual 3D objects) would be involved, because there is no need, trees being what they are. For a building it would be more complicated, but there are less of them fortunately!

As for the craters I also agree with you, the graphical processors of the time were not up to the task of depicting them realistically...but why so horrible textures? The Il2 craters do not resemble at all to bomb craters (have you ever seen earth projections going in zigzag like that?) They do no look like craters at all...It is a tad bizarre as there never was any shortage of real crater pictures around...

[X32Wright]

Still, this means at least that the trees will have to be modeled with a TRUNK and then the 4 sided leaves (top area) so that means 4 triangles plus at least another three triangles for the trunk. Considering that it has no 'bounding boxes' for collision detetcion thats still 4x3=12 polygons per tree X thousands. If you make the tree TRUNK as a SINGLE polygon that ORIENTS to the camera this means 5 polygons per tree, thats still alot to render for most graphic cards.


As for the crater effect they just used texture maps mapped to flat polygons for most effect in this game including smoke and clouds, they just either is animated and moved to create an moving effect. The crate effect itself could just be texture serial replacement.

[BENGALtiger]

A more efficient approach:

How about giving the forest area a damage effect like we have in water (and maybe slightly modified variables for a different effect)

Imagine it.

Softly landing on trees will give an effect like getting stuck in the trees, and then slowly going downward (trough trees, breaking them due to weight) until you reach ground (Il2 water already does this). You can maybe bail out of your plane like you can in water.

Smacking into trees at high speed gives you explosion like in water.

Wingtip touching trees will cause that wing to get break off like in water (or edit some properties from what we have now in Il2, to cause damage instead of total break off of the wing or whatever body part that is in contact with trees [virtually Il2 water])

Have sparks and twigs and leaves flying instead of ripples in water, obviously. Hmm I'm trying to imagine the sound from cockpit :)

It seems like the simplest solution to me. Basically forest areas will be like water, with raised altitude.

What do you guys think?

- Bengal

[X32Wright]

Water is treated like the ground (treated Z and Y axis only even with height field displacement) so I think that depends on the vector computations as well as geometry axis displacement. Since trees cannot be treated like a 'single complete surface' (it has X,Y and Z 'depth' dimensions) since it has dimensions, the problem is much more complex with trees if it is desired to have this interactivity.

Don't get me wrong I want to see this too but I am practical too since I know what would BOG down the resources in the game. I would rather have all tehse extra polygons in the plane for detailed damage modeling rather than being used for trees.

[proton45]

My original thought was that the game use two types of tree models....one type of tree model would be used in areas where a aeroplane might predictably collide with it (like at the end of an airstrips). The other type of tree model would be used in thick dense forest areas. The first type of tree model would have a detailed damage model and a high level of visible detail. This type of tree model would be used in a "limited" fashion. The second type of tree model would be simpler and less cpu intense. This tree model would be used to cover large forested areas...

HAS ANYONE studied the trees that are visible in the "BoB SoW" video clip that came on the bonus DVD in the "IL2 1946" box? I wonder if their are any clues in their that might help determine the technique used to model the forests....

I can't help but think about the 3D grass right now....

[BG-09]

A more efficient approach:
Softly landing on trees will give an effect like getting stuck in the trees, and then slowly going downward (trough trees, breaking them due to weight) until you reach ground (Il2 water already does this). You can maybe bail out of your plane like you can in water.

Smacking into trees at high speed gives you explosion like in water.

Wingtip touching trees will cause that wing to get break off like in water (or edit some properties from what we have now in Il2, to cause damage instead of total break off of the wing or whatever body part that is in contact with trees [virtually Il2 water])

Have sparks and twigs and leaves flying instead of ripples in water, obviously. Hmm I'm trying to imagine the sound from cockpit :)


Nice effects You are describing BENGALtiger. I belive that the forests and trees are the most neglected objects in to the current simulator. But with new CPU, and Olegs determination to rule the world of aircraft simulation in to the next 20 years, we will be granted with marvellous aircrafts and the world for them to fly.

Cheers!


Bg-09