Research On Control Of Air Vehicles Based On Sliding Mode And Adaptive Dynamic Programming

In recent years,technologies in aerospace field develop rapidly,and the relevant theories and techniques in this area have become a research hotspot.As the basic requirement of air vehicles,stable control has received extensive attention.In this thesis,we take two typical air vehicles,i.e.,fixed-wing unmanned aerial vehicle(UAV)and reusable launch vehicle(RLV),as the research objects to study and design flight control and reentry attitude control strategy.As complicated dynamic systems,the speeds and attitudes of fixed-wing UAV and RLV change dramatically.There exist strong nonlinear,coupling,and uncertainty in their aerodynamic characteristics.Complex flight environment brings serious external disturbances.Besides,there also exist other control problems in the above air vehicles,such as unmeasurable angular rates and input constraints.These problems should be taken into account when designing the control strategy.We considered the above control problems to carry out corresponding research and the main contents of the thesis include:Firstly,a multivariable finite-time observer-based adaptive-gain sliding mode control scheme is proposed for a fixed-wing UAV subject to the unmeasurable angular rates and unknown matched/unmatched disturbance.A multi-variable finite-time observer is constructed to achieve the estimation values of unknown states.Based on the observer,a novel adaptive dual-layer continuous terminal sliding mode controller and a tracking differentiator-based adaptive Lipschitz continuous sliding mode controller are designed for attitude and airspeed subsystem and ensure that the fixed-wing UAV can track the reference commands in finite time under the influence of unknown disturbances.Secondly,an adaptive dynamic programming-based adaptive-gain sliding mode control scheme for a fixed-wing UAV with unknown matched and unmatched disturbances is proposed.In this strategy,the control laws are divided into two parts,i.e.,sliding mode control laws and nearly optimal control laws.For attitude and airspeed subsystem,the integral sliding manifolds(ISMs)are designed.According to the different issues in two subsystems,two novel adaptive-gain generalized super-twisting algorithms are developed to design sliding mode control laws and eliminate the effects of disturbances.The system trajectories tend to the designed ISMs in finite time.Then,based on the expected equivalent sliding mode dynamics,a modified adaptive dynamic programming(ADP)approach with actor-critic(AC)structure is utilized to generate the nearly optimal control laws and guarantee the nearly optimal performance of sliding mode dynamics.To a certain extent,the proposed strategy can provide strong robustness and reduce the energy consumption of the system.Thirdly,a fixed-time disturbance observer-based nearly optimal control scheme is proposed for RLV subject to model uncertainties,input constraints,and unknown mismatched/matched disturbances.The dynamics of RLV are divided into outer-loop and inner-loop subsystem.A novel adaptive-gain multivariable generalized super-twisting(AMGST)controller is proposed to design the outer-loop controller.Two modified gain-adaptation laws are derived for tuning the control gains of AMGST controller,which weakens chattering effects and handles the influence of model uncertainties and mismatched disturbances efficiently.For inner-loop subsystem,a fixed-time disturbance observer(FTDO)is adopted to estimate the matched disturbances and the time derivative of virtual control input.Incorporated with the FTDO,a nearly optimal controller,which is based on the critic and actor neural networks(NNs),is utilized to generate the approximate optimal control moments satisfying the input constraints.The stability of the system is proved via Lyapunov technique.Finally,a nearly optimal integral sliding-mode reentry control for RLV is investigated in the presence of parameter uncertainties and external disturbances.For secondorder reentry attitude control model,an integral sliding mode surface is designed and the sliding mode control law based on an adaptive multivariable Lipschitz continuous sliding mode algorithm is developed to handle the parameter uncertainties and external disturbances.Thus,the system maintains on the sliding mode surface and the equivalent sliding-mode dynamic is obtained.Then,an ADP method based on the single critic neural network(CNN)is proposed to obtain the nearly optimal control law,which simplifies the control structure.The tracking errors of RLV and the weight estimation errors of CNN are proved to be uniformly ultimately bounded via Lyapunov technique.The designed control strategy can mitigate the levels of chattering amplitude in control law and then reduce the energy consumption of the control system.