In Chapter 2, a plan was specified by a sequence of actions. This was possible because the effect of actions was completely predictable. Throughout most of Part II, a plan was specified as a path, which is a continuous-stage version of the action sequence. Section 8.2.1 introduced plans that are expressed as a function on the state space. This was optional because uncertainty was not explicitly modeled (except perhaps in the initial state).
As a result of unpredictability caused by nature, it is now important to separate the definition of a plan from its execution. The same plan may be executed many times from the same initial state; however, because of nature, different future states will be obtained. This requires the use of feedback in the form of a plan that maps states to actions.