Two Phase Linear Quadratic Gaussian Controller Design for Launch Vehicle Autopilot

Abstract

The achievable payload of a launch vehicle is sensitive to the mass of the terminal stage which injects the payload into orbit, as well as to tracking errors introduced by the autopilot. The system is usually characterized by lightly damped slosh modes with uncertain parameters, whose stabilization is critical, while maintaining the overall performance of the system. Passive slosh stabilization involves introduction of mechanical baffles into the fluid tanks, which impacts the achievable payload mass. Optimal tracking performance along with active slosh stabilization is hence an ideal combination of autopilot design objectives. While optimal control methods are attractive for such systems, the lack of physical insight in the design procedure often limits the usefulness of these methods for launch vehicle autopilot design. A major challenge in stabilizing lightly damped slosh is the necessity to cater to very large mode magnitudes and phase swings exhibited by the system over the range of parameter perturbations. The significant contribution of this paper is the development of a two phase Linear Quadratic Gaussian approach, utilizing the physical insight into the system performance for tuning the weights in the performance index, to satisfy the time domain as well as frequency domain objectives. A robust design which exhibits superior time domain as well as frequency domain performance as compared to the benchmark classical controller is obtained, with near perfect phase stabilization of the lightly damped slosh dynamics.