Agent Function

A function $f_a$ of an Agent $a$ maps from Percept histories to actions: $f_a:{\mathcal{P}}^{\ast }\to \mathcal{A}$ No multitasking (one action at a time). These functions can become very big and thus might not be computational.

An agent function that is computable can be implemented by an Agent Program that runs on a physical Agent Architecture.