LQG-Based Model Predictive Control of the VLS-1 Attitude at Max Dynamic Pressure

Abstract

ABSTRACT This paper presents the main results of a case study on the attitude control of a satellite launch vehicle using a novel model predictive control (MPC). Design and simulations were based on the Brazilian VLS-1 launcher during the maximum dynamic pressure regime. Such a model was characterized by a 16th-order state-space realization with asymmetrical and coupled attitude dynamics, bending modes, non-collocated sensor-actuator, actuator saturation, wind disturbances, and sensor noise. The idea was to recreate, from a control theory perspective, a scenario similar to the 1999 explosion of the VLS-1 V02, exploring the thesis that the intense vibratory regime could have contributed to the failure. It is demonstrated that when non-collocated sensor-actuator or measurement noise is introduced into the simulations, the control-loop performance degrades and stability may be lost with proportional-integral-derivative-based controllers. To overcome such problems, a new MPC technology is proposed, predictive minimum variance control with full state-feedback. The adopted methodology inherits the advantages of linear quadratic Gaussian with loop-transfer recovery and adds the enhancements of MPC to mitigate the effects of time delays. The results were compared to those obtained by applying a functional approximation of the PI-D controller once designed for the real VLS-1 launcher.