Where is all the ai

#1 You're right, built-in AI in Torque is rather basic. TDN has a good article on the basics of AI scripting, and there are resources like AIGuard and Killer Kork to help you get started.

#2 I think that a upgraded variable AI program, that goes beyond the aiguard, parol and the killer kork, would be great. To allow players to bring up a gui to select a loaded datablock, select a weapon and give them the nessasary slots to input. This would be a boon for TGE and TGEA, becuase James brings up a good point, the aiplayer thats in the starter setup needs to have a upgrade option. And other then the static ones available (which are a pain if you have multiple ai types).

#3 Excellent AI in TGE/A would be ideal if it was added as stock, whether it's for ground or air. It would certainly make the engines more appealing to people looking for high class AI which is required for all good games. Consisting of Pathfinding, attack, defend, find shortest or longest route, pick up objects and drop objects at end destination.

Are there any plans to add more ground breaking AI or even physics to stock TGE/A ?

#4 I just downloaded the quake 3 source and took a look at their code and i see that torque is way behind on the ai part.to put a good ai in the torque gaame engine will take more work than people think and i think this is why they did not add that to the tge.i think people that want to make a good ai need to download the quake 3 source and look at their code and they will lern from that.

#5 AI Is specific to the game you're working on. I'm making an RPG, so i sure don't need my monsters taking cover and trying to hide from me while coordinating assualts from behind.

That said, it's what you make of it. Torque isn't a FPS game engine, it's a game engine. So it can't be programmed to do everything for everyone.

There's several posts/resources about implementing ai, from killer kork, to the Inmersive AI something or another.

#6 Thats true Ramen-sama, but the AIplayer even needs to be updated, even to give a player the ability to pop in a aiplayer per level, by default. We use badbot, aiguard and aipatrol, but there are alot of complexities that need to be addressed, wiether a player needs it or not, all games generally require a AI opponent of some kind. And can be improved on. Right now its a bit rediculous that the aiplayer doesnt tell the difference between levels. I think that simply having a improved AIguard/patrol setup would be fine for alot of folks. but right now, ai that is "standard" is essentially just a pathed character with no behaviours. And the check out the posts for ai, i can probably be safe in saying hundreds of posts. And what is the first thing folks say. Go get AIguard and Ai patrol. Now mind you killer kork and their is a few other ai (such as a aicar, which would be nice for the racing example) Speaking of which, wouldnt that make sense if your making a racing example, to have at least 1 ai for cars as a standard? You have to remember that the examples definitly go beyond a fps example, and folks that and i will qoute you raman

" Torque isn't a FPS game engine, it's a game engine"

So dont you think that ai needs to be revisited as a standard, becuase the current ai could easily be replaced as a standard with the 2 mentioned. That would give folks a example to play with. Even TGB has behavious for object as a standard. Its time we requested some items that are a standard in 99% of engines. We arent talking about creating a huge new class, were talking something thats minor. Most of us that familiar with the engine dont need to have this conversation. But for folks that are looking for answers and solutions. Ai needs to be improved on a basic level. Even if its not used by a player for their game. The engine is shipped with a FPS and a Racing Example that both do not have a ai thats viable for gameplay. Why? Becuase and ill qoute the famous line

Torque isn't a FPS game engine, it's a game engine. So it can't be programmed to do everything for everyone.

No it cant. But for those that want smarter AI, have Immersive AI, Killer Kork, and a number of other "smarter" options including the Quake 3 option mentioned. Just becuase it isnt in the engine doesnt mean folks have to use it. But as a standard aigaurd and a racing ai, should be placed in and implimented for folks to learn the ropes rather then having to go through the loop holes and come here and try to filter through the same answer. If they dont need AI(thats a game i gotta see btw)then they dont place any sample Kork orcs. Just like all the other items "extras" that are in the engine. In todays game market having a AI option is almost a standing must have in a game engine. In Fact, I am very sure that if someone created a improved Ai Pack, it would be a hot seller. Becuase frankly most of the options are pretty cumbersome from a programmers point of view.

#7 GarageGames has a lot on their plate. I would much rather have them spend time upgrading core engine components, than adding in a game specific feature like AI.

#8 And i would agree with that they have alot to do, but i think its not a far stretch that they would start including a slightly improved ai as a "core component" to compete with todays engine market. And to say that AI isnt a core component, thats like saying shaders are a core component too, its not, it just makes it look pretty. Which seems to be what folks want then a fully rounded engine. Though also phyics i suppose is another area that isnt a core issue while we are on that. But most folks would say, physics, shaders and ai are what drive alot of games. And before you say ai isnt a core issue, show me a typical game today that doesnt have ai. In fact, if you have noticed one of the most common areas of complaint by even the high end games. Is that the AI is dumb. Why, becuase we have this idea in our head that ai isnt a core fundamental. It should be, but we put it in a "its to hard" or "its a FPS thing only" and dismiss it, 90% of games (dismissing most puzzlers and brain games) require ai that is greater then pathed. So after AI, id like to see some phyics improvement, There are a few bug fixes, but why do i need to find a bug fix for velocity, when it should be a standard. Yes, GG has alot on its plate, Yes they are doing a great job, the 1.7 beta is the best yet. We love the community response. But simply saying something has a higher proirity doesnt mean a issue shouldnt be addressed.

#9

Quote:But simply saying something has a higher proirity doesnt mean a issue shouldnt be addressed

That is exactly what it means. In a perfect world, Torque would come with state of the art AI for FPS, MMO, RTS, space, adventure games. When you have limited resources you need to complete the higher priority items before the lower.

Maybe a few years down the road when more serious issues are corrected, AI will get another look.

#10 A few years? I do hope, that there is a better outlook for that then a few years. Or is that your opinion of GG in general, that things will take a few years to progress the point where ai will be important to get another look. That seems a bit odd considering the work you yourself im sure have put towards AI in your own game. You dont think maybe starting a mini upgrade of ai would be considerate, regardless of the genre, it needs a upgrade as a standard. Even if its the inclusion of AIgaurd and CarBot as a standard. And im not talking about creating a perfect situational ai for every type. As i said before. TGB has behaviour for objects, TGE and TGEA need a upgrade for the general release of products. And before you say, its not hard to impliment it , no its not, but having folks having to look for a better ai from the get go for a basic functioning ai. shouldnt happen. If your going to have a FPS starter.kit, or a racing starter.kit, you might think these would have something to kick start the ai. Same as if you buy a mmokit, you have at least some basic ai. Let me put it more simply, when i can move a piece of Interior via the pathed interiors tutorial better then i can move my basic Ai with aigaurd, then i think there is a small issue.

#11 AI is NOT easy.

If it was they would just make it all for you and give it to you for free. But making AI for a racing game, or an RTS, or a FPS, are all completely different, and they are all a lot of work.

I would love it if torque came with some sort-of AI framework in place, but I actually have no idea what that would look like and still remain "generic".

Its possible GG will do some updates to the stock AI, but "cutting-edge-AI" is not listed as a feature of Torque, and it isn't their responsibility to provide it.

#12 Nod, we use aiguard, aipatrol and badbots for our ai needs for fantasci. So far its done us well. Just gotta get it working at its full potential.

#13 Why put the shoulder of burden on GG? AAA+ 100k engines aside, do any of the cheaper engines provide a more fleshed out AI? If not then I'm inclined to say AI is our problem.

I came into developing with Torque as a hobby and with no previous game design experience. I've come to learn that yes, developing AI for your game is made much more difficult by the lack of good examples. But it is far from impossible. I've cobbled together a nice A* pathfinding system with job queues for my AI from things I've found posted here. With some luck I'll muddle through path steering and interior node mapping to finish this sucker off. But that just seems like my job, not GGs.

Instead I'd like to see continued developement for larger terrains, opengl support, and rendering enhancements over game-dependant items like AI - the core engine components that far exceed my skill level in developing.

#14 I think your guys are looking past what im saying, andsince when is larger terrains core support, same with opengl since when is that a core item, its a "i would like to see item" not a we must have item. If so You would see opengl for TGEa from day one. and im not talking about popping in some wholy perfect ai, read my post. Im talking about having a basic ai setup, that doesnt duplicate itself every level, that a player might be able to pop in with a gui like everything else. Anyway. The point being, you guys want flash over function, cool with me.

#15 What exactly is the improvement you are requesting?

#16 I didnt request it, but mearly responding to the first post. But it seems i have taken up the torch for this "request".
Well for starters, actually adding in a AIguard/patrol ai would be far more versitile then the pathed ai. The current ai system, is antiquted and anyone actaully needing a ai automatilly gets refered the the aigaurd and aipatrol or killer kork anyway. Even if you do setup a ai, it only currently functions on 1 mission loading up each time. This in and of itself, should be corrected, unless your entire game as a general game engine is going to be 1 mission. I doubt that. So thats 1 huge point, another is, how many pathed ais, do you know, outside of puzzle games? But as its been clearly pointed out, this is a general game engine. So were supposed to use only pathed ai, and add it in as a afterthought, when 90% of the games are actual in need of at least a simple, place and mountable ai. I think alot of folks are looking past the obvious item here. Where there has been some nice improvements. Tosay that GG couldnt make that a standard isnt a hard thing to do, if people ask for it. currently they come to the boards and the community tells them the same answer. goto the same set of ai. Aigaurd, aipatrol, killer Kork and immersive ai. Which leads most to say,hey this sucks, lets impliment our own ai becuase the one that is in there is not even close to what we need. So i would clearly say that the aiguard or even badbots would at the very least be a minimal standard. Same applies to the racing game. I dont know many solo racing games. And even a basic carbot ai, to be implimented would save alot of folks some time to progress forward in those fields. Again, for the record, i have ai and i haveimplimented badbots and aigaurd in my games, and use them well. But for new folks. It should be mandatory as a basic. Aiplayer by itself is a very good class, but having a bitmore meat, like being able to set a behavior or do more then run a next path script. Not asking for a whole lot, just a booster to this area. hundreds of sites offering models, not very many offering ai and programming packs. Ai may be hard. But why make everyone start off, with the same question... um where is all the ai, to which every post responds...

#17 If anything you are being overly generous to the "stock" ai, at best it illustrates how to generate an ai-move. It is just a placeholder and definitely not intended as anything but the barest bones starting point of an AI.

It sounds to me like you are suggesting some sortof behavior architecture. Maybe something like the immersiveAI resource, which defines a goal-based architecture, which you can potentially then make game-specific behaviors for. Sounds great to me.

Then again there are a ton of different AI architectures that could fit for any one game. Maybe you just need some simple FSM(s), maybe its a bit more complex and you need some heirarchical FSM(s), maybe to get the apparence of strategic-thinking you really want you need some kindof planner, etc etc etc...

But there plenty of little things they could do, without a lot of effort, to improve the existing AI -- since theres nothing actually there. But, really, whats the point? Anything they add may or may not apply to any individual making a game, yet it increases the "appearance" that GG IS trying to provide AI for EVERYONE.

As soon as GG puts some AI into the engine, even if it as simple as integrating and fixing up a few resources, that makes those features their responsibility. If it doesn't work exactly the way someone wants, they will probably make tons of flame posts about how GG needs to fix a,b,c and add x,y,z.

It would be awful nice of them to do everything you mentioned, but that makes it look more and more like they are actually providing working AI for your FPS, racer, etc. And they definitely can't do that.

I have a feeling that the rather generic "Torque" engine will see a lot of variation in the future. Maybe theres a whole version of the engine dedicated to racing games, and a whole version dedicated to RTS, etc... When that day comes, I would fully expect it to include some AI appropriate for that genre.

#18 Just remember folks, that while the nifty graphics and terrain stuff is used by almost every game type, AI is genre specific.

Not only is it genre specific, but AI is as challenging as the rest of the game engine. Even A* has a million little variations for different uses. What Edward and James Hawn are really looking for is an open source or low budget superclient. I know of a team that is working on such a thing for Multiverse (FrequencyPattern on the MV forums). Their superclient is probably MV specific, but if people show enough interest, perhaps they could be talked into making it engine agnostic.

Then again, that is MMO AI that they are working on and not terribly useful to things like FPS or RTS games.

#19 "and since when is larger terrains core support, same with opengl since when is that a core item"

Game engines always show the supported render paths in their comparison charts. That is just about as core as you can get. AI can be written entirely in Torquescript whereas altering the render path can require some pretty hefty reworking of the engine.

Larger terrain support is a bit "core" because of the low level MMX hardware calls being made to support the lightmap generation. Again, unlike AI the terrain manager is entirely engine based and a bit more universal then AI. Though I admit the newer TGEA engines are handling it better and some game genres don't need terrain.

In the end I feel like I could sit down and have a working AI in a few week using nothing more then Torquescript, and a pretty robust AI coded in a few months with some small engine modifications. Reworking core engine components like the ones I've mentioned above would take longer and might just be beyond my skill level entirely.

#20 The core reason why you will not see "heavy" AI as part of any of the Torque engines is because no matter how we approach it, it won't be generic enough to qualify for being part of a game engine.

As many have said, each and every game requires different types of AI, and different implementations for optimization. We could implement a basic FSM, and some people will need a goal based solution. We could implement full 3D dynamic A* pathfinding, and RTS games will need an optimized non-dynamic solution.

Of course, you may in the future see free, low cost, or even high cost AI solutions as 3rd party add ons/starter kits, or even GG developed starter kits, but they will aim at specific genres or even specific game implementations.