金慶的專(zhuān)欄

行為樹(shù)的兩種運(yùn)行方式

(金慶的專(zhuān)欄 2017.8)

行為樹(shù)每個(gè)Tick運(yùn)行有兩種實(shí)現(xiàn)方式：
* 從根節(jié)點(diǎn)重新開(kāi)始運(yùn)行
* 繼續(xù)上次運(yùn)行的節(jié)點(diǎn)

http://blog.renatopp.com/2014/08/15/an-introduction-to-behavior-trees-part-3/

> One common question when implementing a Behavior Tree is that: what to do in the next tick after a node returned a running state? There are two answer to it: starting the graph traversal from the running node or starting it over from the first node.

Behavior3 是每次都從根節(jié)點(diǎn)運(yùn)行的。這種實(shí)現(xiàn)簡(jiǎn)單，正在運(yùn)行中的行為可以自然地中止切換到其他行為。

缺點(diǎn)是每次運(yùn)行都是遍歷整個(gè)樹(shù)，對(duì)于龐大的行為樹(shù)，性能較差。

http://www.gamasutra.com/blogs/ChrisSimpson/20140717/221339/Behavior_trees_for_AI_How_they_work.php

> In the basic implementation of behaviour trees, the system will traverse down from the root of the tree every single frame, testing each node down the tree to see which is active, rechecking any nodes along the way, until it reaches the currently active node to tick it again.

> This isn’t a very efficient way to do things, especially when the behaviour tree gets deeper as its developed and expanded during development. I’d say its a must that any behaviour tree you implement should store any currently processing nodes so they can be ticked directly within the behaviour tree engine rather than per tick traversal of the entire tree.

Behaviac 實(shí)現(xiàn)為繼續(xù)當(dāng)前節(jié)點(diǎn)運(yùn)行。

http://www.behaviac.com/concepts/

> 當(dāng)節(jié)點(diǎn)持續(xù)返回“運(yùn)行”的時(shí)候，BT樹(shù)的內(nèi)部“知道”該節(jié)點(diǎn)是在持續(xù)“運(yùn)行”的，從而在后續(xù)的執(zhí)行過(guò)程中“直接”繼續(xù)執(zhí)行該節(jié)點(diǎn)，而不需要從頭開(kāi)始執(zhí)行，直到該運(yùn)行狀態(tài)的節(jié)點(diǎn)返回“成功”或“失敗”，從而繼續(xù)后續(xù)的節(jié)點(diǎn)。

但是為了處理事件打斷當(dāng)前運(yùn)行，需要復(fù)雜的實(shí)現(xiàn)。

> 如果發(fā)生了其他“重要”的事情需要處理怎么辦？
> 在behaviac里至少有多種辦法。

具體為：

* 前置節(jié)點(diǎn)
* 并行節(jié)點(diǎn)
* 監(jiān)測(cè)節(jié)點(diǎn)
* 事件子樹(shù)