游戏AI - 行为树

Question

你如何为游戏制作强大的AI /脚本系统？

1)对于所有NPC的/环境/实体,你给他们一个单独的行为树(等patrolBehavior,allyBehavior,vendorBehavior,doorBehavior)？如果屏幕上有500个单元,我应该在树上进行完整传递(从root - > node/action)还是应该对所有单元进行单节点进度？

2)我在update()函数中做了AI逻辑......但我听说有些游戏有自己独立的AI线程,有什么想法吗？

3)我想知道如何将我的游戏分成几个部分/章节......我是否使用一个简单的变量(EVENT ="Mission 3")来表示玩家的表现如何,并使其全部呈线性？然后利用上面树上的变量？

Answer 1

我会尽力回答你的问题.

1. 我在静态类中执行所有分支逻辑和行为树,例如:

    public static class Behavior
   {
       //Branch
       public static Action Selector(Func<bool> cond, Action ifTrue, Action ifFalse) {
           return () => { if (cond()) { ifTrue(); } else { ifFalse(); } };
           }
       public static Action Sequencer(Action a, Action b) {
             return () => { a(); b(); }
           }
   
       //Example trees
       public static Func<bool> ifPlayerIsInSight = () => { return true; /*...true iff WorldState shows guard can see player...*/};
   
       public static Action shootAtPlayer = () => { /*...aim guard's weapon at player and fire...*/ };
   
       public static Func<bool> ifUnderFire = () => { return true; /*...true iff WorldState shows guard hears player gunfire...*/};
   
       public static Action takeCover = () => { /*...guard runs for nearest shelter... */};
   
       public static Action walkBackAndForthGuardingDoorway = () => { /*...default guard patrol behaviour...*/ };
   
       public static Action patrollingGuardBehaviour =
         Selector(Behavior.ifPlayerIsInSight, Behavior.shootAtPlayer,
           Selector(Behavior.ifUnderFire, Behavior.takeCover,
             Behavior.walkBackAndForthGuardingDoorway));
   }
   

   1. 在LateUpdate()或last中执行此操作,因此它不会滞后主循环.

   2. 它是由你决定.您可以在每个行为树中实现"状态",或者将其拆分并管理在哪个时间使用.