Real-time Physics? Options

[Mole]

While reading various things the other day, I stumbled across this article:

http://news.yahoo.com/s/ap/20060620/ap_on_..._game_physics_2

It is about realtime physics taking games to the next level. I have to say that I agree with it.

I am curious to hear what the rest of you think.

-Mole

[Yellow]

I think its great to have extra power for physics, it would bring a new dimension to the gaming industry. I think it will be slow in spreading though, since few gamers would see the benifit of it just yet, or be willing to spend money on it. But in a few years, it could be quite a hit.

[Mole]

Introducing real-time physics into games is going to be the next "big thing" to come along. Computers are naturally getting more powerful, and these type of effects will be the deciding factor for gamers to choose one title over another.

I am curious to see if people think that this is the type of thing that we should be looking to add into PW or not. I have my personal opinions on it, but I am curious to hear what others think about it.

From the beginning, PW has always tried to make our game cutting edge. Is this the type of cutting edge technology that we should be pursuing?

-Mole

[Yellow]

Looking through sourceforge I already found a few libraries for it, NXogre, as the name implies, is one specialized for Ogre3D:

http://www.nxogre.org/Main_Page

It might be a nice feature to add, but we must be sure that people without the PhysX card would still be able to play and the world stays consistend. With those draw backs, I'm not sure if its worth it. Because it would be twice the work setting up two psysics systems and the card could only be used for gameplay influencial stuff that is also supported by the software version aswell.

[Mole]

There are a bunch of physics engines that we could use. The ones that I can think of off the top of my head are the Open Dynamics Engine, OPAL, Havox, NovodeX, Newton, True Axis, and Tokamak and you already mentioned NxOgre.

You are right, if we are going to implement something like real-world physics in PW, then we have to make sure whatever we implement is compatible with everyone. I do not think that limiting ourselves to supporting a single hardware accelerator would be the right idea, but maybe it is. Whatever we do, the game still has to be just as playable, and look just as good, for those people without the physics card.

Maybe a better question would be this: if we implemented real-world physics, what would we use it for? I would like to hear the PW community weigh in on their thoughts on what we could do by implementing this.

It has to be something more than just making the monsters fall over â??realisticallyâ?� when they are killed.

What do people think?

-Mole

[ghedipunk]

The most obvious uses would be character and object motion... We couldn't use it for determining if projectiles hit their mark, though, because we'd have to calculate the odds server side to prevent exploits... and we don't want to weigh down the server with calculating trajectories.

One thing that a physics engine could do server side is validate how a player moves, making sure that he's not cheating, and moving small objects such as a pile of rocks. Other than that, most physics should be done client side for visual effects... good old falling monsters and such. ;)

[Honis]

I feel that the things that are controlled by a physics chip shouldn't have a bairing on the game play.

Ex: You have 2 people (X and Y) playing in the same area. X has the $5000 computer that is god like. Y has a $500 computer from Gateway. In the game say X blows up a pile of rocks. This results in bebbles, dust, etc. that needs to be calculated. X sees all this happening and says "Coool!!!" Y sees the pile of rocks and then it disappears (probably can support a explosion type transition), his text window informs him X destroyed a pile of rocks. I feel that the resulting pebbles and etc. shouldn't have anything to do with the game. They can be further blown up an manipulated, but all Y sees is X shotting the ground.

Y can also blow up the pile of rocks, but it will result in the same graphic as X blowing it up. The pile of rocks can be manipulated other ways (harvesting, weathering [^_^]) but again, all Y sees is the pile get smaller. X sees dust fly up from his pick hitting it.

The idea of including the physics would be cool, but to keep server lag and hacking down the results of the physics shouldn't affect the world outside of a high end client.

I hope this made sence. I was Y in EQ2 and it restricted what I could see in game. I controlled a clay figure and I found it very stupid since I had full settings in FFXI and it ran and looked better on the same computer.

[Minthos]

Let me tell you this: Oblivion has quite nice physics, but when stuff dies in that game they suddenly get handed over to the physics engine and it looks utterly weird.

What I would like to see physics used for in our game is destroying buildings and walls, and destroying or moving other obstacles (boulders, trees, etc). We could have a threshold for the amont of force required to affect something, and a threshold for the minimum size of something for it to be affected, to avoid excessive calculations.

Server calculates, using a simplified physics formula, what the effects of the change is to gameplay (I can only think of pathfinding right now), and clients draw some fancy graphics depending on their capabilities (using hardware physics, software physics or very simplified physics), to illustrate to the player what the server already knows.

I played a first-person shooter a few years ago that allowed you to destroy parts of the game world (earth, walls, rock, you name it) with explosives, I don't remember what its name was, but I really liked that feature. It wasn't anything fancy, you just saw an explosion and pieces of wall flying everywhere, and then there was a neat little hole in the wall large enough to crawl through. It also had some huge vehicles that could dig through rock. I've never seen any other game do this, perhaps with the exception of a mechwarrior game where you could make shallow craters in the ground with explosives, but that doesen't really count.

In my opinion we should definitely add physics to our todo list. With a bit of clever mathematics I'm sure we can do basic physics in software without too much cpu load, we just have to simplify the formulas as much as possible and leave out some details.

[Mole]

This is kind of a long post and includes replies to a few people... Bear with me here!

[blockquote]Ghedi said:

The most obvious uses would be character and object motion... We couldn't use it for determining if projectiles hit their mark, though, because we'd have to calculate the odds server side to prevent exploits... and we don't want to weigh down the server with calculating trajectories.[/blockquote]

I wonder if we could use the physics engine to determine whether or not a projectile hit its mark. I have not looked at any of the physics engine code so I donâ??t know if this is possible or not. What if we could use the server to calculate the physics part of the trajectory and then passed the calculation off to the client to render it?

I know you said that we donâ??t want to weigh down the server with calculating trajectories, but we could always make a physics server that can do the calculations. The server marshaler could send the physics calculations to the physics server. If the physics server doesnâ??t exist (because we shut it off or whatever) then the client â??calculatesâ?� the physics (which is to not calculate any physics). Either way the projectile gets displayed, one just a little more accurately than the other.

With the distributed nature of the server, having various processes to calculate things would be easy to do.

[blockquote]Ghedi said:

One thing that a physics engine could do server side is validate how a player moves, making sure that he's not cheating,[/blockquote]

Wouldnâ??t this be calculated by the pathfinding algorithms that would calculate how a player moves and to make sure they are moving â??correctlyâ?�?

[blockquote]Ghedi said:

Other than that, most physics should be done client side for visual effects... good old falling monsters and such[/blockquote]

If this is the only thing that we are using a physics engine for, then I would say to not do it. Falling monsters can all be approximated with a simple algorithm. It may not be very realistic, but it can easily be made to look just like every other game in existence and it wouldnâ??t take nearly any processing power to do it.

[blockquote]Honis said:

Ex: You have 2 people (X and Y) playing in the same area. X has the $5000 computer that is god like. Y has a $500 computer from Gateway. In the game say X blows up a pile of rocks. This results in bebbles, dust, etc. that needs to be calculated. X sees all this happening and says "Coool!!!" Y sees the pile of rocks and then it disappears (probably can support a explosion type transition),[/blockquote]

Are you saying that this would be a problem? Or are you saying that if we did it this way, then we could have physics for the people who have the processing power and not have physics for the people who donâ??t have the processing power? I think that if we were to implement physics, then we would have to make sure the people without enough processing power could play the game and still have as much fun.

[blockquote]Minthos said:

What I would like to see physics used for in our game is destroying buildings and walls, and destroying or moving other obstacles (boulders, trees, etc)[/blockquote]

I had an idea to do something like this and it was to allow players to actually cut down trees and have them fall down. This would have the effect of changing the landscape of the game too, and that could be a problem. What would happen if everyone cut down the trees? Would the world run out of trees for a limited time? That might present a whole raft of different problems.

-Mole

[Honis]

Ex: You have 2 people (X and Y) playing in the same area. X has the $5000 computer that is god like. Y has a $500 computer from Gateway. In the game say X blows up a pile of rocks. This results in bebbles, dust, etc. that needs to be calculated. X sees all this happening and says "Coool!!!" Y sees the pile of rocks and then it disappears (probably can support a explosion type transition),

Are you saying that this would be a problem? Or are you saying that if we did it this way, then we could have physics for the people who have the processing power and not have physics for the people who donâ??t have the processing power? I think that if we were to implement physics, then we would have to make sure the people without enough processing power could play the game and still have as much fun.

I'm just trying to avoid a transition difference. When I played FFXI it looked good and most computers could play it all full settings because it uses simple texturing and use of the same models just dressed differently.

When I played EQ2 the graphics made every character including myself look like clay figures. The buildings looked like blocks and I was still lagging everywhere.

I feel this is the direction the physics engines are going. You'll play a game that looks and runs great on your compy. Then you get a game that has the same primise (MMOs its lvl and collect stuff, FPS its shoot the thing that moves, etc.) but this extra physics capability. It'll look great on a high end, but most people don't have that. So, you have those that don't want to spend $XXX on physic's going back to what looks and runs better on their machine for free.

I would like the physics to just help it look prettier and not effect any gameplay. This will let someone turn it off and keep the game looking pretty.

[Minthos]

[quote1151472209=Mole][blockquote]Minthos said:

What I would like to see physics used for in our game is destroying buildings and walls, and destroying or moving other obstacles (boulders, trees, etc)[/blockquote]
I had an idea to do something like this and it was to allow players to actually cut down trees and have them fall down. This would have the effect of changing the landscape of the game too, and that could be a problem. What would happen if everyone cut down the trees? Would the world run out of trees for a limited time? That might present a whole raft of different problems.

-Mole
[/quote1151472209]
No no, it would present opportunities!

We'd have to make sure that altering the landscape is a huge task, and that the landscape (at least vegetation) restores itself at a rather rapid pace. Forest fires would be fun and all, but it might be just a tad too devastating to our game world and tempt players to keep our game world a barren wasteland permanently

[Minthos]

And since you asked for more ideas in irc, here's one idea. I'll describe a scenario to illustrate.

A group of players wants to decorate their fortress with trees. They chop down the tree they want to move to their fortress, wait for it to respawn as a small bush, then they bring a shovel and a wheelbarrow and move the bush to where they want it, and plant it there. Simple enough, done by a single player in a few days. I figure a tree respawns after one to three days, takes about a week to reach a size where it could be called small tree, and a month to reach full size.

Okay, so what if they for some reason want to move a small tree, without having to wait for it to respawn? They would need a small cart to move it (ox or horse cart), they would need several people to uproot it and lift it onto the cart. Now, how would the act of lifting and moving a tree performed by multiple players be handled by the game? Here's where the physics engine comes into play. They have a button for "grab", and when they have grabbed the tree they could have one button for lift and one for lower, and one for toss. The players could synchronize their actions on teamspeak/ventrilo, so they all lift at the same time and toss the tree onto the cart at the same time.

What if they want to move something even bigger? Something so big they can barely lift it? Then they would need a ramp to get it onto the cart. In real life such a ramp could perhaps be constructed by placing some sturdy planks from the ground onto the cart. With a physics engine, we could let players do that in the game as well.

[Mole]

[quote1151515862=Honis]
I would like the physics to just help it look prettier and not effect any gameplay. This will let someone turn it off and keep the game looking pretty.
[/quote1151515862]

I could not have said this better myself. Everything that PW does in our game MUST be this way. The game must be playable to everyone on a "modern day computer". We will have to draw the line somewhere, because it would not be in our best interest to support people who want to play on an old 486 or something. Lower limits will have to be set, but I do not ses a problem with that. What I do have a problem with is setting our lower limits too high.

Any real-time physics that are implemented must not detract from the game if they are turned off by the user. Turning them on must only add to the experience.

-Mole

[Mole]

[quote1151516331=Minthos]
No no, it would present opportunities!

We'd have to make sure that altering the landscape is a huge task, and that the landscape (at least vegetation) restores itself at a rather rapid pace. Forest fires would be fun and all, but it might be just a tad too devastating to our game world and tempt players to keep our game world a barren wasteland permanently
[/quote1151516331]

I agree that it would be very interesting to implement. I also think that it would be an MMO first if we did something like this. (Has any other MMO done something like this before?)

There are a few problems that I see regarding it though that keep making me talk myself out of the idea every time.

You named one of them. What would keep players from trying to deforest the world? Would this be a good thing?

How grpahically and server intensive would it be to do this? We would basically have to keep track of every tree. There would be millions of them. Is it really worth the effort to do this?

How much art work would we need here?

No matter what we did, we could still implement forest fires. We would just have to do it on a much smaller scale. Clients would see the change, and then the forest would we out of resources.

-Mole

[Minthos]

[quote1151560620=Mole]We would basically have to keep track of every tree. There would be millions of them.[/quote1151560620]
How many millions?

Let's say each tree has a pair(do we need triplet?) of 32-bit coordinates, that's 8(12) bytes. It has an address to its art, that's 4 more bytes. It has a size, that's 1 byte, it has another byte for scorched, a third byte for chopped.. that's 15(19) bytes, let's bump it up to 16(20) for alignment reasons(I'm sure we can come up with something to use that extra space for), add another 4 bytes for its address, then we have 20(24) bytes per tree.

That's 20(24) megabytes of server memory per 1 million trees.

The client would only have to keep track of each individual tree in its vincinity. For trees farther away, the server could just tell the client "all trees in area between x1,y1 and x2,y2 are (insert correct value between 0 - 255) scorched", and client would draw that area as brown/gray/black/orange/yellow (depending on the scorched value) instead of green.

Yes, it would be more work for us. I think it would be worth it :)

[Minthos]

[quote1151562206=Mole]What would keep players from trying to deforest the world? Would this be a good thing?
[/quote1151562206]
Regarding forest fires.. Honesty I have not a single clue.

Apart from forest fires, there is one good way of preventing deforestation, and that's making sure the pace at which trees respawn and regrow is higher than the pace at which our target number of players could theoretically manage to chop them down if they were all online 24/7 and did nothing but chop trees.

[Mole]

[blockquote]Minthos said:
Apart from forest fires, there is one good way of preventing deforestation, and that's making sure the pace at which trees respawn and regrow is higher than the pace at which our target number of players could theoretically manage to chop them down if they were all online 24/7 and did nothing but chop trees.[/blockquote]

Hmmm. That would all depend on how many players we had and how big the world was. I think something like that would force limits on ourselves that we may not want. With only 10 players, keeping the trees alive is easy. With 10,000 it gets harder. And if we hand a million players, that would mean our landmass would be huge.

It's a good idea, but I think imposing limits on our game like this might come back to haunt us in the future.

-Mole

[ghedipunk]

For trees specifically, make saplings undesirable... The faster that an area is deforested, and unless replaced with farms, then the faster saplings grow... Good soil + more sunlight = quick restocking.

[Minthos]

[quote1151654246=Mole]
Hmmm. That would all depend on how many players we had and how big the world was. I think something like that would force limits on ourselves that we may not want. With only 10 players, keeping the trees alive is easy. With 10,000 it gets harder. And if we hand a million players, that would mean our landmass would be huge.

It's a good idea, but I think imposing limits on our game like this might come back to haunt us in the future.[/quote1151654246]
There's a limit to how many players we can realistically cram into an area of the world without crowding it, and the way I see it we should only open a small part of the world initially, and open up larger areas as we get more players and as our artists create those areas. If our estimate regarding how fast they can cut down the trees turns out to be way off, we can always increase the regrowth rate and/or add more trees.

[RicoSuave]

Yes, but that's what all the BETAs are for... stress testing and determining what needs to be tweaked. And you're right, I suspect the admins will continually tweak to keep not just content alive, but also the ambiance.