A classic AI bot dilemma
Written by MM. Posted at 8:42 am on September 9th, 2009
7 comments.
Post a comment.
Status Updates
- Hey, we searched all over the database, but we couldn't find any status updates. Have you posted any yet?
aha, so that’s what you mean. something similar:
http://i30.tinypic.com/mmcv8o.jpg
true, all areas are reachable, but that doesn’t help a bot figure out how to get from point a to point b.
oh, and link-dead drawing copyright http://mm.soldat.pl
Are bots such a priority?
I mean no matter how clever bots you make people will still prefer to play against living players.
Of course if you want to make the most awesome bot in the world for your own fun I can understand that but they really are just fillers in a multiplayer game.
You can simple use randon priority to choose the path =)
I’m a bot AI newbie, but what about (for the above scenario, and niko’s) calculating a few moves ahead, to see which one reaches point B?
You could also give bots a preference to always jump (safer) to the location they want to get to based upon their current goal list.
Decomposition is the key here. How does ‘Karel’ react to the given obstacles and what does it choose?
Sorry you drawing just reminds me of Karel learns Java. I love that little robot.
Are bots such a priority?
I need to test gameplay with bots. I can’t be sitting all day long with people.
I think that you should look at a bot as a human player. When you spawn in DM in Soldat your ultimate goal is to find and kill an enemy. Now it’s up to you to choose the right path. Every player is focusing on the map’s “hot spots” and usually chooses to go there as fast as he can so as far you can make a long jump somewhere – you do it. You should take a look at Unreal Engine 3 feature which is generating paths and hotspots on the fly so designers can adjust the maps to be perfect. Your path finding algorithm could be very useful with it. Maybe generating 4 spawn points on the map and making the bots find all the possible movements, choosing the fastest, not necessarily shortest paths and I bet that there will be places in which they cross and that’s how you should be able to find the hotspots on every map. Now, every time a bot spawns he knows where he can most likely find someone to kill. With objective-based gameplay it’s more complicated. But bots should have their own goals, so some are defending the main objective, some are attacking, some are supporting and some are picking up incoming enemies as snipers or placing traps (whatever you plan to include). These hotspots may be very helpful in choosing their right paths. Once it is clear the most important thing would be combat behaviour. I think that you did it pretty well in Soldat, except bots shooting like fokkin morons. They also shoot have priorities and the main one in combat should be maintaining a position higher then opponents and finishing him off(bots in Soldat where only interested in you as long they saw you – no memory, no goals, no foreseeing ). I can’t imagine how awesome would it be to see bots that can not only focus their attention on you, find you and even outsmart you like throwing a grenade to the other end of a tunnel cuz you might escape from there etc.
I think that you should keep on working on bots but making them as good as I think you want to may postpone the game itself another year or two.
Good luck.