A FAM-based Agent for a Ball and Beam

Abstract

This paper presents a reactive agent built under the Subsumption Architecture, which purpose is to control a Ball and Beam system. The architecture of the agent is implemented making use of a fuzzy associative memory (FAM) and feedback control laws. The FAM is used to address the action selection problem known for this architecture. The FAM defines different fuzzy conditions. These conditions are associated to feedback control laws. The integrated control signal is obtained by defuzzifying the FAM through the weighted average technique. The time-response of the agent is simulated on a model of the Ball and Beam system and validated with an implementation at the laboratory. The results show the behavior of the variables of the system and the dynamic characteristics of the agent. The control signals stay active at every moment and the FAM allows smooth transitions between them, depending on the fuzzy condition of the system.