Page 202 - Deep Learning
P. 202

The Growth of Competence                 185

               Table 6.1.  Goal-situation-action rules that implement the plan in Figure 6.2.

            Goal            Situation                  Actions
            Go to Nth floor via   Standing by elevator door &  Set goal to “Call elevator”
             the elevator     elevator is at some other floor &
                              call button is not lit
            Call elevator   Standing by elevator door &  push call button
                              elevator is at some other floor &
                              call button is not lit
            Call elevator   Standing by elevator door &  Wait
                              elevator is at some other floor &
                              call button is lit
            Call elevator   Standing by elevator door &  Enter elevator; goal is sat-
                              elevator is at the current floor &  isfied; set goal to “Use
                              elevator door is open      elevator”
            Use elevator    is inside elevator & floor button   Set goal to “Select floor”
                              is not lit
            Select floor    is inside elevator &       push button for Nth floor;
                              button for Nth floor is not lit   goal is satisfied; set goal to
                              & button is within reach   “Ride to Nth floor”
            Select floor    is inside elevator &       Ask a person to push the
                              button for Nth floor is not lit   button; goal is satisfied; set
                              & button is not within reach  goal to “Ride to Nth floor”
            Ride to floor   is inside elevator & elevator is   Wait
                              moving & button for Nth floor
                              is lit
            Ride to floor   is inside elevator & elevator is at   Exit the elevator; goal is
                              the Nth floor & elevator door   satisfied
                              is open
            Go to Nth floor via   is on Nth floor & outside elevator   Goal is satisfied
             the elevator



            action of the selected rule performed. Each turn through this perceive- decide-
            act cycle results in some change in the current situation, and the cycle repeats.
            A task performance normally consists of a long sequence of perceive-decide-
            act cycles, each cycle ending with either a motor action or a cognitive operation
            such as an inference or the setting of a subgoal. in programming terminology,
            execution is iterative.
               Because  practical  knowledge  is  hierarchically  organized,  execution  is
            also recursive. That is, the top goal is unpacked into a set of subgoals and the
            latter are processed and unpacked in turn by the same processes. Eventually,
            the process of unpacking encounters elementary actions, which are carried
   197   198   199   200   201   202   203   204   205   206   207