Modelagem e controle de um aeropêndulo por meio de redes neurais lagrangianas.

Abstract

This work proposes: the determination of an analytical model and a control law for the airpendulum system through Lagrangian mechanics; The determination of a data-driven model of the aeropendulum using Lagrangian Neural Networks. To carry out this work, concepts such as networks neural networks, Lagrangian mechanics, Newtonian mechanics, numerical integration, symbolic mathematics and analog control. The behavior of the aeropendulum was modeled in three diferent ways: using Newtons and Kirchhofs laws, using Lagrangian mechanics and using Lagrangian Neural Networks. Validation of analytical solutions was carried out using Newtons and Kirchhofs laws and using Lagrangian mechanics. Furthermore, using the Energy Shaping technique, a controller for the angular position of the aeropendulum was designed and validated.