Artificial Intelligence


“Scripting AI” and behavior trees are the same paradigm. The idea is to execute a hierarchical sequence of actions and this results into intelligent behavior. One option in overcome some problems of behavior trees is to separate the individual actions and their effects. This allows to recombine an existing behavior tree into a new sequence of tasks.[1]

A more elaborated way is to a use dedicated Action description language for formalizing temporal logic. This is called a GOAP planner.[2] In tradition of “Shakey the robot”, an abstract goal is formulated and the robot is searching for a symbolic path towards this goal. The disadvantage of GOAP is, that before the planner can schedule the tasks, he needs a world model which is formalized in the action language. This world model provides the preconditions and effects of each action.


  1. Martens, Chris, Eric Butler, and Joseph C. Osborn. "A Resourceful Reframing of Behavior Trees." arXiv preprint arXiv:1803.09099 (2018).
  2. Sharma, Nitish. Reactive synthesis of action planners. Diss. 2019.