Mobile robot navigation with generalized modus ponens type fuzzy reasoning

Abstract

When mobile robots have imprecise sensory information, Mamdani and Sugeno type heuristic procedures are inapplicable to such situations. The fuzzy inference rule known as the generalized modus ponens (GMP), on the other hand, is the only logic based fuzzy inference mechanism which can be applied to combine imprecise observations with the linguistic rules of the rule base and determine an appropriate course of action. It has not been used in real-time applications mainly due to its computational requirements, especially with multi-antecedent rules, and in part due to a lack of knowledge about relationships between various operators used in the GMP. To deal with the complexity issue, operation decomposition is added. The complexity is thus reduced from an exponential function to a polynomial function of the number of antecedent variables. A unified and generalized framework is proposed to study and analyze the relationship among various operators used in GMP. Desirable properties of inference are used as the compatibility criteria in operator selection. Operator combinations satisfying these criteria are identified and referred to as compatible sets of operators.