A Novel Attitude Controller for Hypersonic Aircraft Based on FR-PI2

Due to the uncertainties and various noises of the Hypersonic Aircraft(HA), traditional method of designing attitude controllers based on accurate models faces immense challenges. To solve those proposed urgent problems effectively, in this paper, a innovative data-driven RL algorithm called FR-PI2 is proposed for the task of Hypersonic Aircraft attitude control, wherein several key techniques are introduced. Firstly, Hypersonic Aircraft attitude control is transformed into a stochastic optimal control problem, and the numerical solutions of the proposed stochastic optimal control problem are obtained with a path-integral-based method, called Policy Improvement with Path Integrals(PI2). Secondly, a model-free policy parameterization method with PID controller is integrated into the PI2 algorithm to ensure a low dependency of the Hypersonic Aircraft system model. Thirdly, a brand-new sample filtration method is proposed to guarantee rapid convergence, and an online optimization technology called rolling optimization is applied to guarantee optimality and real-time performance. Two series of experiments are carried out and analyzed to demonstrate the outstanding performance of the FR-PI2 algorithm.