Formalización de controladores difusos.

Abstract

The document presents an extensive analysis of the LAMDA method, an artificial intelligence tool that can work on supervised and unsupervised learning tasks. Initially, the study focuses on identifying the weak points of the algorithm to propose extensions that improve its performance. We propose LAMDA-HAD for classification and LAMDA-RD for clustering. These proposals have been formalized and validated with different benchmarks that have made it possible to study the performance of the proposed extensions and their comparative analysis. In a second stage, we formalize LAMDA in control systems, taking advantage of its ability to detect functional states. For LAMDA to function as a controller, it requires a phase of defining classes that are the functional states of the system and defining fuzzy rules to obtain the control action. Furthermore, as it is a classification algorithm, in this work an inference method is formalized that allows the controller's output to be calculated. Finally, since LSMC requires the knowledge of the expert for the design, an adaptive LAMDA scheme has been proposed. The novelty of this proposal is that for the first time LAMDA is used for fuzzy modeling and control of complex systems. LAMDA controllers have been validated in different case studies, such as in the control of industrial processes with variable dynamics or in the field of robotics for the control of trajectory tracking, the results of which have been analyzed and compared with other similar control techniques, demonstrating that the proposed methods improve overall system performance.