Game Elements : Facial Expression and Game AI

[spmisra]

Hi Guys,

I have put up this post to get comments/suggestion/ideas on the Facial Expression and also the AI system of the game. This will give us a good idea and pathway also for long term. I guess these are also some components that help put-up a great game.
First of all about facial expression
As we will be enhancing the game we will be in need of facial expression component to the game. This will add more realism to the game and also better feeling.
I know of the librarie
http://expression.sourceforge.net/

For AI we will have to go for custom AI. Initially we can start with steering behavior. Then advance gradually as the game keeps on evolving.

 

[TorakunSama]

If we're including facial expressions, we might as well go and add language, maybe a text-to-voice with lips synch. But as well as I understand this frst step, faces will not much be animated, as they might not be visible, unless Jochen decides to allow avatars. And then again, there might not be any avatar to avatar interaction...
But I like the idea, for the main game. As for AI, creatures, ships, drones, missiles and mines, wont really need facial expressions  . Or did I misunderstood you?

[spmisra]

TorakunSama: Both are different components. I added them both as when going for some good game, I guess we will need them.
For that we need to build up the engine. All of us are working part-time on this engine as of now, so adding brick by brick will be good. To get some acceptable performance from them it will take some good amount of time.
For Drones, creatures etc we can go for some basic AI. No facial expression is needed for anything accept for human.
For lipsync we do not need text to voice. From the wav file we can do a basic lipsync.
http://jlipsync.lamhauge.dk/
JLipSync is a program for creating timesheets/x-sheets/exposure sheets to sync animation to a pre-recorded audio track
It calculates the rise and fall of voice and then arranges lips according to it.
It may not be exact but its good.
http://www.meloware.com/blender/lipsync.htm
This is also a free software written in Python.
http://www.lostmarble.com/papagayo/index.shtml

But right now my mail problem is I am not able to get the editor up also . Please help me in it.

[Jochen Stier]

But right now my mail problem is I am not able to get the editor up also . Please help me in it.

I have decided not to support the open source release I made a while ago, because it will just be too much work. There are so many different versions of Windows and Visual Studio in circulation that I could be chasing individual link errors forever. There are still some ways I can think of for software developers to contribute.

One way is to study the source release and develop your add-on components such as AI, speech etc. in such a way that it will be easy to integrate in the future. When your extension is complete and robust enough we will integrate it via a dll or something like that. This is the most risky though as I am most likely to develop the most effective AI myself. Simply because of the knowledge I have of Geist3D in terms of collision detection, ray casting etc

Another way is to write custom generators for some of the artifacts in Geist3D, such as clouds or tree distribution. The clouds for example are pre-computed and stored in a file containing the particles make up individual clouds, cloud fields and cluster. Whatever data I will come up with, will be quite rudimentary and I am sure it will be possible to improve upon the individual clouds and their distribution quite a bit. I will open up the file format and then anyone can just write their custom generators. All you have to do is replace the corresponding file in the models directory and voila, your clouds are in the environment. I think the same can be done for the trees. In the future this will be the preferred way for other software developers to extent the engine. This way the interfaces are clear and there is little in terms of me debugging someone else code, or someone else debugging mine.

The third form is in terms of using the Lua scripting interface. Of course this is not as powerful as it only contains Geist3D specific artifacts. But, it will important for level design and user interface in the future. I will put a high priority on supporting any bug fixes/feature request regarding the Lua Api.

I will release the clouds file format as soon as I can followed by a file format for the trees. Stay tuned…

[spmisra]

Jochen: I don't to start up this thread out here, but just for a little brief. I am able to build everything clean, the only thing I get is during execution. The error is posted in the technical help section. This may be due to my graphic card, or something. Please have a look at it when you get chance.

[TorakunSama]

OK people, we are falling behind! The year is almost over, let's pick up the pace!

I believe is time to start discussing about AI strategy.
I think AI pilots at first must be able to go for a check point and avoid obstacle on the way there. The intriging part of it might be when the AI must decide if it must go up, down, left or right!! maybe depending on space.
My idea was to have an invisible projection of a ship's boundary and attach a collision sensor to it. If it collides the AI must try another angle. So it rotates in the path ahead swiping for the best path to take.

Ok now, this is just a simplistic way of thinking result of blender exposure 

[Jochen Stier]

This is actually not a bad idea at all!

[TorakunSama]

Really?
The same can be applied for the players, we add a display for a first person camera, representing the ships size. That way players will know it the ship fits in-between those rocks or not.
For AI, the best path would be the center of the checkpoint. Since most races will be aerial, the checkpoints will determine the track's course. As the ships fly, aiming for the center of the checkpoint,they will dodge objects according to the ship's position. If the ship is below in relationship with the checkpoint, and the obstacle is just in front, then the AI will pick an higher path, if there's room (the ship itself must have proximity sensors). If there's no room, the ship must slow down till is clear or accelerate to get there first, the last being the more efficient, if the ship is a fast one...
This might work even for AI limited by a 2D transform abilities (stuck on the ground e.i. animals, bots, vehicles).