Method RT (ms) ST 5% criteria (ms) ST 2% criteria (ms) Backstepping 1.4 1.6 1.81 P&O 1.3 59 NA FLBC 1.3 1.9 NA Method SSE (%) Overshoot (V) Ripples (V) Backstepping 0.16 0.6 0.9 P&O 0.3 70 8.9 FLBC 0.29 12.8 9
Fuzzy logic-based controllers (FLBCs) or algorithms incorporate the human knowledge and information of a particular system in determining a fuzzy rule base to control it.
The state of the art for FLBC systems is best described as being in the exploratory phase.
We are exploring the systematic use of a full-fledged FLBC (see McCarthy [1982] for an early vetting of this idea).
Having presented the pertinent essentials of SAT (see Kimbrough and Lee [1986] and Moore [1993] for additional information), we shall now discuss a specific application area, an Army office environment, for application of an FLBC.
We chose an Army office environment to test the application of the FLBC approach because we are familiar with it and because the clear lines of authority in an Army office present opportunities for computerized inferencing on messages.
We give further analysis, specific to our FLBC and our prototype implementation, in Section 5.
We now consider how to represent the seven message types, discussed in Section 4.2, in an FLBC. Although we shall develop a particular language, we hypothesize that the family of languages to which it belongs, FLBC-2 (see below), is in fact quite general and can be applied in many contexts besides the particular application we are presently reporting on.
The form of an FLBC message can be summarized with the definition shown in Figure 1, which defines a family of languages.
The FLBC representation of a statement message type is the following:
Let interesting (H, I) belong to our FLBC lexicon with the intended interpretation that the information item named by I is interesting to person H.
Our method is simply to treat a query as its own illocutionary force and to place the knowledge of what to do in response to a question in the programs that use and process the FLBC messages.