View Full Version : a taste of AI programming

02-03-2005, 01:40 PM
I wonder how many have ever played Carnage Heart on the playstation? If you did then how good was you in programming your OKE (mechs)?

Carnage Heart is a very hardcore strategy game(yes even consoles have hardcore games)where you design and build you mechs then program them. After you program them you can see how well they fight in real time againest other AI mechs. Programing the AI (which is the heart of the game)has been simpified by placing computer chips on a board which the program runs. Many of these chips has varible which you can adjust. (Example: range) there is a total of 40 chips all together. While you will learn to build a AI in no time yet building a good one will be the hard part.
It hard enough making a good AI to fight one on one but most battles in the game will be 3 vs 3. The trick is to program your mechs to work as a team. At first it didn't seem to hard but it didn't take long to see a AI I thought was good (well atleast by itself) doing totally stupid things. One program my Mechs was doing nothing but hopping around like grasshoppers (they was dodging too much)while another program had one mech stand still while another friendly mech was shooting him in the back. I tried to fix it so to avoid firing on friendly only to see the one in the back do nothing while watching his friend being gunned down. I had them run into each other or all of them too bunched together or others too far apart. Most of the time It's better to start a new AI program than trying to fix the old one. I could also build flying OKE or tanks which requirs a whole new program.
After I've played Carnage Heart, even though I'm not a AI programmer, I have some idea how tough it is to program an AI even in something as simple as avoiding crashing into a hill or crashing into each other, as well as having the AI work in a group. If programming IL2/PF AI is as tough as programming mechs in CH , Then so far I'm impress. If you like to trying AI programming and have a PS/PS2 then give Carnage Heart a try.
P.S Too bad it's not online. You can only test your AI mechs againest a friends by memory card.

02-03-2005, 03:19 PM

Well, I've done some programming... And still do every once in a while.
Not using any form of AI though.
I think it's already a very difficult job to try and predict what a true and real Human can input and push buttons and be ready to deal with that!

Indeed, programming AI is a dounting task, and although there seem to be flaws here, I too am impressed in what it is capable of ...

02-03-2005, 06:00 PM
Interesting read, thanks. The ultimate goal of Artificial Intelligence research is to make AI behave the same as a humanoid. I don't think that goal can ever be reached, but it can be approximated better and better perhaps.
<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Most of the time It's better to start a new AI program than trying to fix the old one. <HR></BLOCKQUOTE>
We know Oleg tweaks and twists our AI, for better or worse, but he said he will be doing totally new things for BoB.

02-03-2005, 06:43 PM
I don't know. Maybe I have. What's a Playstation?

02-03-2005, 07:06 PM
i had to programme a game in c++ for my uni some time ago.

it was very simple, nothing major, but i found to get the bad guys to move around in anything other than a random manner was to give them a memory.

i would design my player object to have description fields: as in, blue eyes, black coat, beard etc.

when the gane would load, i would have info on the player for the bad guys, or they had to start fresh.

as the bad guys saw the player, depending on the distance meant they could see my black coat but not my blue eyes... so they made a choice based on the info they had whether or not to close in and check. If i escaped they could either wait there for me to come back, or make a decision to go parallel to me on the city block and keep tabs on me. ONce they decided i didnt have a gun they would close in.

what it boiled down to was that i had an immense amount of info to drive just one bad guy, or many if they all acted like clones.

it took ages to get it right, and build up a case for "reasoned" decision making rather than a course of action determined by a random number generator.

in these days of limited resources ie time, refining, money, i think that theyve done very well.

those knockers who like to run game AI down, should give it a try themselves.... http://forums.ubi.com/images/smilies/88.gif

02-04-2005, 07:15 AM
<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>The ultimate goal of Artificial Intelligence research is to make AI behave the same as a humanoid. <HR></BLOCKQUOTE>
No, the goal of AI is to create non-organic sentience. It does not have to be in the form of a humanoid. If it is ever acheived, it will undoubtedly take on many forms. This can range from insects to Generals that plan all our army's strategic moves for us (like Whopper in war games). Humanoid AI (the classical robot) is undoubtedly the most romanticized, perhaps due to a subconscious desire to create in our own image? In any case, true AI will take some time to achieve. One of the biggest problems with AI is deciding when something actually obtains intelligence and or sentience. That subject has been the subject of philosophical debate for many centuries.