Abstraction as a Way of Uncertainty Representation in Smart Rules Engine

When creating the Internet of Things and Smart Machines, the RULES ENGINE pattern is developed. The tasks that Rules Engine solves in real time are characterized by large amounts of data from various sensors. To account for uncertainty, a new model of Smart Rules Engine is proposed with a two-step inference engine, including an engine of abstraction from data. A model based on quantitative, definitive and generalization type of abstraction is considered. It is shown that the proposed model solves the problem of the bridge between natural language, as a carrier of the concept's meaning, and data from sensors, as carriers of the perception. A model for presenting a prototype is given in which incomplete knowledge, expert's uncertainty in the definition of the concept and aging of data are taken into account. The prototype and data model representation and the abstraction engine are built based on fuzzy L-R numbers and arithmetic operations on them. The preliminary step of abstraction from the data, besides taking into account the uncertainty of a different nature, significantly reduces the dimension of the fuzzy inference space. A comparison of the two-step fuzzy inference engine with the traditional inference engine Fuzzy Logic System Type-1 is given using the example of a wheeled robot making the decision to safely cross an unregulated crossroad.