An architecture for intelligent sensors and fuzzy inputs for fuzzy logic controllers

Abstract

Many papers have shown that fuzzy logic can be successfully applied to problems that are nonlinear in nature. Specifically in the area of control, many fuzzy logic controllers (FLCS) have been shown to be excellent means of control, most notably in situations where an adequate model of the system is not available. But the controller may only be as good as the quality of the sensor values that provide input to it. The quality of the sensor values is of great concern for designers. Just as in the cases where an accurate model of the system is not easily obtained, if at all, fuzzy controllers with fuzzy inputs are examined for their potential to compensate for lack of information about the distribution of the noise in the inputs. An architecture for an "intelligent" sensor is presented. The architecture encompasses sensor filtering, sensor fusion and anomaly detection that can be used with any controller to improve sensor values. The research presented here focuses on the portion of the architecture that is responsible for taking the crisp inputs and fuzzifying them. Tests of the architecture are performed with a FLC as the underlying controller since FLCs can readily deal with fuzzified inputs without modification (ie. defuzzified). Results show that the fuzzy inputs may: (1) improve performance in situations where noise is present, (2) provide the means of compensating for unavoidable delays in systems, and (3) maintain reasonable performance measures with less sampling. A simulation and a real world application of the intelligent sensor architecture are presented for analysis.