Sunday, July 14, 2013
Summary
There's a very interesting paradigm shift that's happening right now in the AI. This is something that I'm both excited and nervous about, as it's going to be different from the traditional model of AI in RTSs. In Limit Theory, when you tell your unit to do something, you will not be saying "go attack this ship." You will not be injecting an order into the unit. You will simply be saying "this ship needs to be destroyed." You will be suggesting to your unit that you have a desire for something to happen. This is a very fundamental difference, and I know that I've discussed it as early as January of this year (or maybe it was December of last year?)
Since then, however, I reverted to a more traditional AI approach of simply issuing orders. This is how LTP works and it's how basically every other game works. But now that actions are gone and events are in place, I'm being led back in the other direction
How does it affect your gameplay experience? Very deeply. In a perfect world where the AI works well, it will allow you to trust your units far more than you would in a traditional game. It may even eliminate micromanagement. For example, suppose you select a group of units scattered across a system, and tell them to attack an enemy ship. In a traditional game you would get all of your units heading toward the ship and firing on it when they got close enough. They would probably even use the "fastest path," so that would mean using warp lanes as necessary. That's what you would expect to happen when you say "I need you to attack this ship right now."
But what if you said "I need this ship to be destroyed?" Imagine that some of your ships are damaged and that a friendly station is in the system. Your AI recognizes that it does not have nearly as high a chance of successfully doing what you want without first repairing itself. So your units dock at the station and purchase a repair. But the AI might also recognize that it has a much higher chance of succeeding if it stays together as a group. So the non-damaged units wait outside the station, and when the repairs are finished, the whole fleet congregates on a common ground before launching an attack. Heck, maybe the AI could even hire mercenary help if it recognizes that the chances of success are low
This probably sounds scary to people. It sounds scary to me. But imagine a world in which you trust your AI that much. You trust that when you say "I need this station to be destroyed," you can simply leave it at that and move on to your other tasks, rather than worrying about everything. If you have competent pilots, they'll do it just as well as if you had sat there and micromanaged the whole thing. Maybe you could configure your AI extensively to control the level of autonomy that it is allowed (for example, whether your pilots are allowed to buy repairs, ammunition, hire help, etc.)
Just some food for thought that's brought about by this change in the AI system. I make no promises that the LT AI will be that smart, but the neat thing is that it is a natural possibility with this type of AI
[ You can visit devtime.ltheory.com for more detailed information on today's work. ]
Post
Mon Jul 15, 2013 12:21 am
#1
Week of July 14, 2013
“Whether you think you can, or you think you can't--you're right.” ~ Henry Ford