AI and you

Hello again,

today you’re going to help us lay the foundation for AoA’s survivor and enemy AI, but more on that in a moment. AI is one of those oft-neglected, yet vitally important aspects of video games. Making smart AI that can challenge the player is hard and too often developers skimp out on it by padding enemy numbers and statistics instead. This can create a predictable and boring experience for players.

There are many formal approaches to game design. One of them revolves around the concept of “stories.” Now this isn’t necessarily the kind of story your grandma used to tell you. Rather, a story is a short (1-2 sentences) description of a game event and the consequences thereof, as seen from the eyes of the player.

For example: “As you kill the enemy leader, their group cohesion degrades” is a story. It tells of a player’s interaction with a game mechanic (killing an enemy leader) and the consequence (they no longer fight as cohesively). This allows us to decide what kind of experiences we want to deliver in our game and lets us derive technical requirements. From this example we know, we need enemies to fight as part of a defined group. Such groups must have their own dedicated AI which comes in smarter and dumber variants and depends on the amount of bullet holes you put in a specific actor.

In practice this could look like this: your expedition team attacks a group of raiders guarding several crates of loot. As most of the raiders rush into cover, one of them hangs back and repeatedly shouts at the others. One raider switches his assault rifle to full auto and starts laying down suppressive fire as others move to flank you. When he’s out of ammo another raider takes over and covers his reload.

Recognizing the enemy leader in the back, you order your sniper to target him. As he goes down the raiders’ cohesion drops. Some fall back to the nearest position of cover and start shooting at the nearest target, others rush forward blindly. Without covering fire from their friends, the rushing raiders are easy pickings and soon you’re suppressing and flanking the remainders yourself.

Another example: “The hungry space wolf attacks your survivor, but flees when shot in the leg.” We now know we want our AI to make decisions based on their current needs, but also be able to re-evaluate situations and prioritize higher goals such as self-preservation.

“After chopping down the last player-designated tree, the survivor heads back for lunch but drops the newly cut timber off at storage on the way.” Like the earlier example, our AI needs to fulfill it’s needs, but for player-controlled survivors it must also prioritize work assignments. It must also be able to recognize opportunities for work optimization and perform tasks that are on the way to its next goal.

And where do you play into this? Simple: we want you guys to send us your best AI stories. Tell us your ideas for smart raider tactics, clever worker time-savers and animal behaviors. This way you help us figure out our AI requirements to make sure we deliver to you the best game possible. Let us know in the comments or join our Discord and tell us directly!