Reentry Maneuvering Control For Hypersonic Vehicles Based On Dynamic Structure Functional Link Network

Considering the increasingly tense international situation and the further requirement of national defense security, the hypersonic vehicle(HSV) has become a research focus of world military powers, and it has immeasurable strategic value in the military and civilian fields. The study of maneuver control system is always the important part of the flight control system of the HSV. However, the control of HSV is confronted with the unknown dynamical disturbance and parameter uncertainty problem during its reentry phase. This paper focuses on the research of following aspects:Firstly, we refer to the related literatures at home and abroad, then use the mathematical model in earth-centered inertial frame, presenting the deducing process in brief. Its aim is to improve the accuracy of the model to some extent.Secondly, considering the effect of unknown uncertainty and disturbance during the reentry, a control scheme which combines the baseline nonlinear generalized predictive control(NGPC) controller with the functional link network(FLN) control compensator.Thirdly, in view of the defects of the fixed structure network, a new dynamic structure FLN(DSFLN) is presented to optimize the FLN control compensator. The Lyapunov function is utilized to grow the DSFLN, which combines the network scale with the system performance effectively, and the pruning strategy is put forward according to output features of the FLN nodes, which can increase the system real-time performance. The online training algorithm of DSFLN weights is derived from Lyapunov theory. The simulation results show that the DSFLN-based controller can find the appropriate network topology to approximate the disturbances/uncertainties during reentry and compensate for them to achieve good interference rejection capability and satisfactory control performance.Finally, an improved DSFLN which contains the feedback information is put forward as a dynamical structure recurrent FLN(DSRFLN) method. It unions the DSRFLN with a gain-variable robust item to approximate the disturbance/uncertainty in both track angle loop and attitude loop. The simulation results show that the method can track the instruction signals and accomplish the lateral maneuver.