We review two fuzzy-logic approaches for the control of systems operating in highly uncertain environments. These systems are characterized by requirements that, going beyond specific control needs, demand the ability to react and overcome unexpected circumstances. In a typical application, the underlying controller must, therefore, seek several explicitly stated objectives while reacting to various unexpected circumstances. Our techniques are based on the utilization of special conceptual structures and tools for the definition and implementation of hierarchical supervisory controllers that seek to attain, to the best possible extent, both purposive and reactive goals. The first type of methods is based on the application of rule-based metalevel control techniques to determine the nature of the possibility distributions that describe the relative overall desirability of potential control actions. Lower-level control is achieved by means of special control structures that seek to induce behaviors aimed at specific goals or groups of goals. High-level controllers determine, on the basis of deliberations about the current operational context, the tradeoffs that must be made to attain the goals of lower level controllers. The second class of techniques-aimed at problems where fast response is essential-are based on an axiomatic approach to the definition of context-dependent acceptability of controls. Our discussion focuses primarily upon formal aspects of these methods related to the analytical determination of the properties of controllers and controlled systems.
展开▼
