ATTITUDE CONTROLLER DESIGN OF THE BRAZILIAN SATELLITE LAUNCHER VIA HYBRID NEURAL-GENETIC APPROACH

Resumo

This work proposes the design of an attitude controller for the Brazilian launching vehicle via
mode-selection using a hybrid neural-genetic method. Given the high complexity of the rocket dynamic
equations, the model was linearised and minimized with a model order reduction technique, in particular
mode-selection. The hybrid approach performs the weighting matrices search of the linear quadratic
(LQ) method and the solution of the Algebraic Riccati Equation (ARE) that leads to the attitude controller
gains. The performance analysis of the reduced order model and the designed controller was performed
in the frequency and time domain, while the hybrid neural-genetic approach was evaluated through fitness
function and energy and infinity norms, respectively. The proposed controller reached the time domain
specifications, i.e. rise time, settling time and overshoot for the maximum dynamic pressure instant. The
results suggest that the hybrid approach could speed up the attitude controller design process of Brazilian
launchers, reducing costs and re-design possibility.