基于神经网络观测器的水下拖体输出反馈姿态控制

针对水下拖体俯仰姿态控制系统设计时无法精确建模、模型具有强烈非线性和不确定性以及不可测外界干扰和角速度等问题，设计了一种基于观测器的补偿控制系统。该系统中两个神经网络独立对动态非线性模型进行在线辨识，分别配合状态观测器和补偿滑模控制器完成对系统状态变量的实时观测与补偿控制。在观测器中加入鲁棒项抑制附加干扰，设计了两种自适应权值更新律以及一种映射修正自适应律以保证系统的稳定性。经Lyapunov理论证明，在满足一定条件下，系统的观测与控制误差均为最终一致有界。仿真和实验结果表明，所设计的补偿控制系统具有良好的自适应性和鲁棒性。

Neural Network Observer-based Output Attitude Control of a Towed Underwater Vehicle

The attitude control of towed underwater vehicles (TUVs) is a challenging problem due to the strong nonlinearity and uncertainty of dynamic model, and unmeasurable external disturbances and angular velocity. An observer-based output compensation controller was designed for a TUV, in which two neural networks are employed to identify the dynamic nonlinear model on-line for the state observer and compensated sliding-mode controller, respectively. And a robust term is added to the observer to suppress additional interference. The adaptive weight updating laws and a projection modification adaptive law are designed for neural networks to ensure the stability of the system. It is proved by the Lyapunov method that the observation and control errors of the system are uniformly and ultimately bounded under certain conditions. Simulated and experimental results show that the designed compensation control system has good adaptability and robustness.