Observer-based backstepping longitudinal control for carrier-based UAV with actuator faults

The paper presents the longitudinal control for the carrier-based unmanned aerial vehicle (UAV) system with unmeasured states, actuator faults, control input constraints, and external disturbances. By combining output state observer, adaptive fuzzy control, and constraint backstepping technology, a robust fault tolerant control approach is proposed. An output state observer with fuzzy logic systems is developed to estimate unmeasured states, and command filters rather than differentiations of virtual control law are used to solve the computational complexity problem in traditional backstepping. Additionally, a robust term is introduced to offset the fuzzy adaptive estimation error and external disturbance, and an appropriate fault controller structure with matching conditions obtained from fault compensation is proposed. Based on the Lyapunov theory, the designed control program is illustrated to guarantee that all the closed-loop signals of the given system are bounded, and the output errors converge to a small neighborhood of zero. A carrier-based UAV nonlinear longitudinal model is employed to testify the feasibility and validity of the control scheme. The simulation results show that all the controllers can perform at a satisfactory level of reference tracking despite the existence of unknown aerodynamic parameters and actuator faults. © 2017 Beijing Institute of Aerospace Information.

航天飞行器控制系统误差实时补偿方案研究

箭上计算机实时计算因惯性器件系统误差而造成的测速误差，并将其引入导航方程后，通过导引和关机进行实时补偿。数学仿真表明采用这种补偿方案，箭载计算机的计算工作量小，补偿精度高，是一种有实用价值的方案。

棒材连轧机立式活套的多变量自适应解耦控制

将非线性补偿和 PI控制策略相结合 ,针对棒材连轧机立式活套系统提出了一种多变量自适应解耦控制器 ,从理论上证明了闭环系统大范围渐近稳定 ,自适应参数能以很小的误差跟踪实际参数的变化 ,并给出了可用于控制器参数整定的系统动态性能与控制器参数间的定量关系 .

基于LSTM与牛顿迭代的两轴系统轮廓误差控制

针对轮廓误差影响运动系统精度的问题,提出结合长短期记忆神经网络(LSTM)和牛顿迭代法对轮廓误差进行预测、通过转换任务坐标系对轮廓误差进行补偿的方法.在运动平台上提取特征轮廓与数据,将牛顿迭代法应用于对轮廓误差的计算,通过计算出的轮廓误差对优化后的LSTM神经网络进行训练,建立更准确的轮廓误差预测模型.通过转换任务坐标系,将预测的轮廓误差作为前馈补偿到参考轮廓中,提高轮廓控制性能.通过试验对比PID、迭代法和神经网络法,利用随机NRBUS轨迹验证泛化性,表明提出的方法能够有效地预测并控制轮廓误差,在精密运动控制领域有良好的应用前景.

基于预测补偿模型的绳索并联机构预松弛控制

本文分析了含绳长误差的绳索并联机构运动控制过程,设计了一种基于长短期记忆(LSTM)神经网络预测误差补偿模型的预松弛控制方法,提高了绳索并联机构末端位姿在运动过程中的准确性与连续性。本文基于矢量闭环原理建立了系统运动学模型,得到了绳索末端位姿的非线性误差模型,采用LSTM神经网络进行非线性误差的预测补偿。基于离散控制周期分配主、从控制绳,实现系统的预松弛控制,减小绳索末端的无序晃动。仿真结果表明,误差补偿之后,末端位姿精度有了明显的提高,而预松弛控制的位姿与索力相较于传统控制也更加连续,验证了该方法的可行性。