Here I just write down thoughts and notes that came up during developing.
Finite State Machines vs. Behavioural Trees
So with FSM you have a dedicated class for one task. For example „Walk to Food“. You then can set the food object by casting the base class state class to the Walk ToFood class and access the member that holds the food object, thus its position.
With Behavioural Tress you need to pass a context, like a dictionary consisting of a Object/Unity.Object and a identifier string.
Not sure yet which is better in which situation.