Hybrid Adaptive Launch Vehicle Ascent Flight Control

Abstract

One of the main challenges facing the designers of launch vehicle ascent flight control systems is dealing with the effects of un-modeled dynamics or operation under off-nominal flight conditions. An adaptive control system would allow the plant dynamics to deviate beyond their nominal design limits while sustaining satisfactory operation. For the purpose of maintaining stability and improving performance in adverse flight conditions, a hybrid adaptive control scheme is selected and implemented as a launch vehicle flight controller. This architecture merges both direct and indirect adaptive elements within a classical dynamic inversion controller. The combination of these elements enables the control system to retain the nonlinear capabilities of an adaptive network while relying heavily on the linear element to dictate the dynamic response under most operating conditions. Initially, the ascent dynamics of a launch vehicle resembling the Crew Launch Vehicle are summarized in both nonlinear and linearized form. Development of the hybrid adaptive launch vehicle flight control architecture, which includes the conversion of guidance command input and progression of the neural network weight update law, is then discussed. To validate the hybrid adaptive flight controller, a high-fidelity launch vehicle ascent flight simulator was obtained from NASA Marshall along with a traditional PID ascent flight controller for performance comparison. Ascent simulations indicate that the hybrid adaptive flight controller is able to reduce guidance tracking errors significantly more than its traditional PID counterpart. However, this reduction in tracking error is coupled with the appearance of small spikes in the engine gimbal commands. Consequently, the avoidance of harmful interactions with structural bending modes needs to be investigated.