1 registered members (henrybane),
1,499
guests, and 1
spider. |
Key:
Admin,
Global Mod,
Mod
|
|
|
Intense Ai: Get the Demo Now! (and screenies too!)
#193834
02/13/05 13:45
02/13/05 13:45
|
|
Joined: Jul 2004
Posts: 1,205 Greece
OP
Serious User
Intense Ai: Get the Demo Now! (and screenies too!)
Hello everyone!
I saved my 100th post for this really special occasion.
The very first demo of Intense Ai is out. Please take the time to download it and tell me what you think of it. Basic features are there, so you can get a hint of how the Basic version of the ai Series will look like.
I apologise in advance for the size of this demo. It's something less than 50mb, but do give it a shot cause it's well worth it. If you find that too much for you though, I guess you'll have to do with some screens instead..
Get the Demo Here!
The features of the Intense Ai engine go as follows:
Nodeless Pathfinding (You don't have to spend hours placing nodes in your levels any more. Ai is capable of handling itself just fine..)
Advanced Cover Tactics (Using the latest technology on cover Tactics, as used in Medal of Honor and Call of Duty. Just build your level and let our Bots worry about where to hide)
Alliances (If a team of gangsters meet up with an FBI squad, hell will break loose! You make the teams, and bots will stick to them!)
Plug and Play (Does C-Script looks like Greek to you? No worries, you can install Intense Ai without typing a single line of Code. If you need to configure your bots properly, you can now do it through WED instead!)
Template 6 Friendly (You don't like the templates? That's fine, our engine will work without them. But if you do, Intense Ai will happily accept any Template 6 weapons you've made already and play along realy nice with your existing Template 6 project.
If you played the demo, and you felt like the models you saw should actually be in your game as well, well I guess this is your lucky day.. Models are on sale as well.
Everything was made by the very talented Blattsalat we all know and love, and he carries a big selection of ww2 models that will be available for sale as well, under the name of 'Intense Ai - ww2 model pack'
The pack will include
Humanoid Models (Soldiers, Generals, and pretty much everything that will make up for a ww2 game)
Weapons (blattsalat has a 'thing' for world war weapons, so you may be assured that everything in this pack will be an EXACT replica of the real thing..)
Heavy and Light artillery (Tanks, ballistics, planes, you name it.. we got it..)
Everything animated and ready to rock 'n roll(Everything is fully animated, and completely compatible with the Intense Ai engine. Either you buy this pack along with the Ai engine or not, these ww2 pack is for you)
Ok, I won't advertise any longer, I feel like these guys on midnight advertisment TV shows who just won't shut up and keep praising their product on and on and on.... until you fall asleep:)
I started the Intense Series, firstly because I could, and secondly because I felt like the community would greatly benefit from it. There are lots of people who have amazing and very original game ideas, but they hardly manage to go through the first stages of game production, because they have to deal with basic things like Ai, pathfinding, etc, etc.. My goal is to build a ground level for them, so that they can take the time to work on their ideas instead!
Of course, I can't just give my work away like that. I have spent many many months on this project, so I'll have to get something back in order to keep working on it. So I want to take this opportunity to ask everyone who is interested in this project to list All the features they would like to see in the Basic version of Intense Ai, and the price they would be willing to pay for it. I do intend to make an Advanced and a Proffesional version as well(I know, the edition names ain't too original, but they do their job..) so please don't storm these thread with too crazy features, like driving Tanks or Bones animation for now. Keep in mind, that the more features you ask for the Basic version, the more time I'll have to take to make them, and the higher the price will be.. So be sensible:)
If you've read through here... thank you:) If you've downloaded the demo as well, double thanks.. Please take the time to post too. Your feedback means lot to me!
Aris
|
|
|
Re: Intense Ai: Get the Demo Now! (and screenies t
[Re: LarryLaffer]
#193856
02/15/05 04:16
02/15/05 04:16
|
Joined: Nov 2003
Posts: 299
profmakx
Member
|
Member
Joined: Nov 2003
Posts: 299
|
Quote:
Hello everyone!
The features of the Intense Ai engine go as follows: Nodeless Pathfinding (You don't have to spend hours placing nodes in your levels any more. Ai is capable of handling itself just fine..)
I would be interested in the way you do this. Is it efficient (also for "real" 3D not only pseudo 2.5D (meaning people running around on the ground). Do the units _plan_ their path or just "find" apath? Iīve done some work on path-planning for units and havent yet thought up an efficient way of doing this (well without nodes, or placing nodes myself or similar)
If you donīt want to answer because itīs a secret of yours or patented, donīt worry. Otherwise you can as well answer my per PM, perhaps we can agree on some technology or money-transfer of some kind
profmakx
|
|
|
Re: Intense Ai: Get the Demo Now! (and screenies t
[Re: profmakx]
#193857
02/15/05 06:48
02/15/05 06:48
|
Joined: Jul 2004
Posts: 1,205 Greece
LarryLaffer
OP
Serious User
|
OP
Serious User
Joined: Jul 2004
Posts: 1,205
Greece
|
Hiya profmakx
I use A* for the Bot pathfinding.. A* prototypes are originally 2d Based(meaning, it might get trickier if your level contains stairs, elevators etc..) but it's still very possible to get in 3d and many many people have a working version of one..
The Bots plan their path, they don't find it.. A* works like this.. The Bot(I'll refer to it as 'Agent' from now on cause it's cooler) creates nodes at runtime toward its destination point. If an obstacle is found in the way, it will try to create the next node in a given angle instead(45o is good) and check if an obstacle is found again. This proccess is repeated, taking into consideration not to create already visited nodes, and always storing information for each node(distance from Agent, Distance from Destination, Parent Node, etc). If a Node makes it to the Destination, a Search through the nodes is done in order to find the shortest path.. That's just Math now...
Here's a simple example shot..
The Red dots are all the nodes created, while the yellow nodes are the ones qualified for the final path.
This Shot only displays the final Path that the Agent will follow..
If there's no path available, you'll be able to see clearer the grid that is being created by the A*
In the example above, I've limited the maximum nodes that can be created for each pathfinding cycle for performance issues. You can see that the nodes ain't enough to test for an opening in the other side of the square.. In the picture below, I've increased the maximum nodes to 1000 instead..
A* can give you great results in most situations and its quite fast too, that's why it's currently the most popular pathfinding technique used in games.. Here's a slightly more complex situation with A* at work..
And here's some more in-game shots as well... I use the new collision system to make the most effecient use of A* with arbitary level geometry, bulk models, etc..
Agent goes to take cover
Agent Goes to kick some butt
Please feel free to PM me and I can give you my pathfinding code for free. But I use nothing more than plain old A*, so no secret formula there..
I really recommend you this book for Pathfinding in Games. AI Game Programming Wisdom I. Also, there's a very fine A* tutorial somewhere around written by Phillip Walser who actually guides you on how to implement A* in GameStudio. And you can also search for LocoWeed's version of Walser's implementation which supports multiple agents running the algorithm at once and considering other models as obstacles as well.
I chose to post here instead of PMing you just in case someone else is also interested in A* Pathfinding. You may also reply here as well, I don't mind turning this thread into an A* discussion, until I work on my product some more..
Cheers and good luck
Aris
|
|
|
Re: Intense Ai: Get the Demo Now! (and screenies t
[Re: LarryLaffer]
#193858
02/15/05 07:39
02/15/05 07:39
|
Joined: Mar 2001
Posts: 3,298 Beverly, Massachusetts
Rhuarc
Expert
|
Expert
Joined: Mar 2001
Posts: 3,298
Beverly, Massachusetts
|
Ahh, this right here is your performance problem, Larry . It is not in the video card, but in the processing power. You are doing an intense amount of pathfinding calculations here, in addition to traces that are very slow. While it probably doesn't seem too bad to calculate all those nodes dynamically, it adds up very quickly; adding the intense tracing is VERY taxing on a system. I'm not certain why, but some systems take traces a lot better than others, even when one is a "more powerful" system. The AI shown here so far is very good, and it's pathfinding IS excellent- it needs a lot more flesh though (jumping, more states, weapon switching, more dynamic movement, etc). Here are my reccommendations to your performance problem, several of these are ones the I myself am using (actually... all of them, LOL): - Use time slicing! If you limit the number of cycles per frame for pathfinding, and divide these cycles evenly among the AI requesting a path, you'll have a much more controllable performance. The problem that comes up here is that a bot may have to wait several frames before it receives its path (or up to seconds in heavy situations). Best way around this is to have the bot use a simple pathfinding state where it will move towards the target using obstacle testing and avoiding nearby obstacles. This is usually not very noticable in game. Another level/method of time slicing is to give for example out of 10 frames(or cycles), 5 frames to pathfinding, 1 to decisionmaking, 3 to targeting and attacks, and 1 more to other needs. This is a bit harder to do because it has to be done by frame, and not by cycles, but it can still be beneficial if done correctly. - At least have support for precompiled paths. If people are using massive levels or want absolute control over paths, an option to precompile the paths and allow them to be read from a static path table is essential. This is a good solution for many people who need to lower the performance cost of pathfinding for lower target sys. requirements. - Use lookups. Rather than using vec_dist every time you check a node, do it once and store the value. This can add up to a big time saver in calculations. Also, lookups for sin/cos when calculating out these node positions would even help. Things like this you need to look for adding support for, even if your system runs by default with dynamic values, allowing a user to define a variable (i.e. "define USE_SINCOS_LOOKUP") to use an alternative performance boost is a BIG plus. Keep in mind, with lookups- you are balancing performance vs. memory usage. If you aren't worried about memory consumption, you can sacrafice some memory in lookups to acheive faster performance, and vice versa. Hope these ideas help . -Rhuarc
|
|
|
Re: Intense Ai: Get the Demo Now! (and screenies t
[Re: Rhuarc]
#193859
02/15/05 08:31
02/15/05 08:31
|
Joined: Jul 2004
Posts: 1,205 Greece
LarryLaffer
OP
Serious User
|
OP
Serious User
Joined: Jul 2004
Posts: 1,205
Greece
|
Hey Rhuarc! No man, the Ai calculations aren't the main cause of the demo's poor performance. Trust me, I've done so many optimizing tests the past week, that I got a fade image of the F11 debug-panel in my vision everywhere I go... I'm sure there are ways to optimise it even further though, and your ideas are gold and I'll keep them in mind.. I think I've already brought it down to a very satisfactory result though, given a reasonable number of Bots on screen... Anyhow, I'll make that cheaper level version of the demo soon, so that will tell us if it's the level or the ai that messes things up.. I'm pretty sure its the level though. Quote:
Use time slicing! If you limit the number of cycles per frame for pathfinding
lol.. come on man, I came up with that idea.. I've said it in that Cover Thread, and you told me it was a pretty good idea.. I've seen you have extended it though..frame limit all of the bot's calculation for better tweaking.. I like, it looks interesting..
Quote:
Best way around this is to have the bot use a simple pathfinding state where it will move towards the target using obstacle testing and avoiding nearby obstacles. This is usually not very noticable in game.
Nuhhh.. Dirty Code... I would avoid junking my pathfinding code like that.. that's all but proffesional. If the bots want to move but they haven't pathfinded their way yet, they stay until they do. They will more probably keep shooting at the player, or keep doing other things. If they need to leave inside an exact time frame though, I'd calculate the average lag it takes for the pathfinding at runtime, then attempt the pathfinding proccess earlier.
Quote:
At least have support for precompiled paths
Well, that kinda ruins the whole idea of not having to do anything to the level in order to run Intense Ai... Lower Systems just use a lighter version of A* with less traces per cycle and less dense grids. I do this already for Low-Med and high versions of the demo so I guess the lighter level version will tell us whether it works or not...
Quote:
Use lookups. Rather than using vec_dist every time you check a node, do it once and store the value
Why don't you like vec_dist? it's a very simple calculation, it's just the square of the difference of the two vectors, powered up and added together, something like that.. I don't think it takes more performance than doing sin and cos calculations and storing values in arrays.
I appreciate your performance ideas. If the Ai still turns out to be heavy I'll go really silly on optimization until I get it right. If no one complains in the next version about performance though, I'll assume its fast enough already.
|
|
|
Re: Intense Ai: Get the Demo Now! (and screenies t
[Re: LarryLaffer]
#193860
02/15/05 10:07
02/15/05 10:07
|
Joined: Mar 2001
Posts: 3,298 Beverly, Massachusetts
Rhuarc
Expert
|
Expert
Joined: Mar 2001
Posts: 3,298
Beverly, Massachusetts
|
Quote:
lol.. come on man, I came up with that idea.. I've said it in that Cover Thread, and you told me it was a pretty good idea.. I've seen you have extended it though..frame limit all of the bot's calculation for better tweaking.. I like, it looks interesting..
LOL... I'd forgotten all about that . I actually was having some issues in a similar manner and ran across the concept in a book.
-Rhuarc
|
|
|
Re: Intense Ai: Get the Demo Now! (and screenies t
[Re: LarryLaffer]
#193862
02/15/05 16:28
02/15/05 16:28
|
Joined: May 2002
Posts: 2,541 Berlin
EX Citer
Expert
|
Expert
Joined: May 2002
Posts: 2,541
Berlin
|
If I could download it I would finish it for sure.(anyway to make the download smaller, like smaller textures?)
:L
|
|
|
|