Fuzzy Logic-Based Intelligent Systems for Uncertainty Handling

Abstract

Uncertainty is intrinsic to real-world intelligent systems, arising from imprecise sensor measurements, incomplete domain knowledge, vague linguistic
descriptions, and stochastic environmental dynamics. Fuzzy logic — providing a formal calculus for graded truth values and linguistic variable
manipulation — remains uniquely suited to modelling and propagating such uncertainty in a mathematically tractable and human-interpretable
manner. This study presents FuzzyCore, a unified fuzzy logic-based intelligent systems framework integrating type-2 fuzzy sets for second-order
uncertainty representation, a genetic algorithm-optimised fuzzy rule base, and a fuzzy cognitive map (FCM)-driven causal reasoning engine for
complex multi-variable decision support. FuzzyCore was evaluated across four uncertainty-rich intelligent system applications: autonomous robot
navigation under sensor noise (ROS2 TurtleBot4, Gaussian noise sigma = 0.15 m), medical diagnosis under incomplete symptom data (UCI Heart
Disease dataset, 32.4% missing values), financial portfolio risk assessment under market volatility regime uncertainty (STOXX Europe 600,
2015–2025), and supply chain disruption risk scoring under geopolitical uncertainty (500 suppliers, 18 risk dimensions). FuzzyCore achieved
navigation success rate of 94.3%, diagnosis AUC-ROC of 0.934, portfolio Sharpe ratio improvement of 0.38, and supply chain risk precision of 0.917
— surpassing crisp rule-based, probabilistic, and type-1 fuzzy baselines by 7.4–18.3%. Interpretability ratings from 36 domain experts averaged
4.5/5.0 — significantly exceeding neural network baselines (2.7/5.0). These results establish FuzzyCore as a robust, interpretable, and practically
deployable framework for uncertainty handling across diverse intelligent system domains.