Hello guys! A while back, I made some postings regarding the AI in a military simulator I'm working on (with two other guys). Some of you requested screenshots etc. but I had none at the time. I now have two posters presenting the project: Poster 1 (There´s a brief description of the 'tactical pathfinding' on this poster) Poster 2 To give you an impression of the simulation, I have some screenshots of the soldier model: Soldier image 1 Soldier image 2 And of the terrain renderer as it is now Terrain picture Impressions, comments or questions are very welcome, especially regarding the AI I posted this because I think it is interesting to show/see the AI in a context, instead of all these isolated A* questions Best regards Ulf Livoff p.s Unfortunately, there is no demo that can be put on the web at this moment. [edited by - UlfLivoff on September 2, 2003 6:15:34 AM]

Is this simulation intended to be realistic (something you''d sell to the military) or merely "believable"?

realistic all the way!

Currently, I''m developing it as a master thesis in cooperation with the danish army. After my graduation in january, it´s my plan to either continue the development with the army, or to continue to develop it as an operation-flashpoint style game or both

Best regards
Ulf

Certainly looking very nice man. It''s a pity that graphics can be shown off in stasis but AI needs a dynamic context to appreciate it.

Mike

Good luck, you have a focused goal, and it will be very interesting. A sh$t load to do if you wanna reach 1.1.2004 though...

By the way, I absolutely loved Operation Flashpoint!!!! (In fact I was kind-of thinking about it as I looked at your posters, remembering Bohemia was also working on a military sim after OF.)

- John

An intresting alternative to A* for short range pathing within a dynamic enviroment is using hardware accelerated voronoi diagrams to generate paths, which appear more natural, looks like it has better awareness of the terrains contour, and can intergrate tactical information more succiently.

Many FPS have bots are superhuman in their ablities, instant head shots from across the map etc... The real trick will be modeling the human reactions and faults realisitcly.

The terrain looks great! Is it using the splatting method?

Good Luck!

-ddn

Thanks for all your comments!

quote:A sh$t load to do if you wanna reach 1.1.2004 though...

Yes I know! Well, our goal for 1.1/04 is primeraly to get all parts of the AI-state machine running, since that actually is the scope of my master thesis.The rest of course, is required to make a game (or simulator) and will be implemented along the way.

quote:By the way, I absolutely loved Operation Flashpoint!!!! (In fact I was kind-of thinking about it as I looked at your posters, remembering Bohemia was also working on a military sim after OF.)

Loved? you mean you dont still play it ? shame on you!
Yes they're a great inspiration (both OFP and the military version "Virtual Battlefield").

quote:
An intresting alternative to A* for short range pathing within a dynamic enviroment is using hardware accelerated voronoi diagrams to generate paths, which appear more natural, looks like it has better awareness of the terrains contour, and can intergrate tactical information more succiently.

Interesting ... I´ll look into that...

quote:
Many FPS have bots are superhuman in their ablities, instant head shots from across the map etc... The real trick will be modeling the human reactions and faults realisitcly.

Actually, I never understood why realistic shooting is so crappy implemented in all games on the market. I mean, get some statistics from a shooting range and implement it

Instant headshots with an uzi from 350 meters is not something you will see in this simulator!

When I was in the army, we learned that a maximum-accuracy shot takes approx. 4 seconds and effective range without optics is approx. 300 meters (for the H&K G3A3 Assault rifle) It´s statistics like that, that I'm going to implement.

Since the manuals I´m using also contains a listing of common errors, I hope to make some realistic non-perfect behaviour without using too much black magic

quote:
The terrain looks great! Is it using the splatting method?

Splatting ??? hmmm never heard that word before ... perhaps because I'm an AI programmer
Actually, I dont know all the bits and pieces of the terrain engine. I mostly posted it to show some context.

If there is more comments or questions, please feel free!

Ulf

[edited by - UlfLivoff on September 2, 2003 5:33:47 PM]

Here ya go :

http://www.cbloom.com/3d/techdocs/splatting.txt

This site descibes the terrain splatting method.

Will the project involve multiusers interacting in realtime with AI bots in a realistic 3D combat arena? If so, what model do you invision for the AI within such an enviroment?

-ddn

The missions will be realistic (e.g. ambushing a convoy) and not CS or BF1942-like.

The AI enemies (Bots) will normally operate in squads of 8 (including the squad leader) with specific objectives assigned to each squad. This means that you''ll never see an AI running around on it''s own spraying bullets in all directions (unless he''s got a serious screw loose).

The AI method for this i Finite state machins to handle the AI''s decisions and a variant of encoded lists to handle it''s planned actions.

So there will be no Unreal tournament-style fighting, if that''s what youre asking for, but yes - It will be a realtime simulator.

Eventually multiplayer will be implemented but for now, it seems a bit far off in the future.

To get the idea of this project: Imagine Microsoft Flight Simulator realism, soldiers instead of airplanes, easier to use and at a slightly higher pace .

Ulf

It all looks very sharp Ulf... I was particularly impressed by the comparison of the photo of the soldier squatting and the rendered version.

I have a couple of thoughts though...

Have you checked out the work done in the Intelligent Agents lab at Melbourne Uni, in particular Emma Norling''s work. She''s also looking at imbuing artificial soldiers with realistic human behaviours and is working in conjunction with Australia''s Defence Science Technology Organisation (the government''s military research fascility). You might benefit from reviewing her work.

Additionally, how do you plan on dealing with uncertainty within your simulation. You''ve mentioned that your AI will be FSM based. That''s not very realistic and real soldiers will very quickly be able to discern an artificial enemy from a human one within the simulation, making the virtual environment less believable.

Cheers,

Timkin