基于单隐层神经网络的空天飞行器鲁棒自适应轨迹线性化控制

Robust Adaptive Trajectory Linearization Control for Aerospace Vehicle Using Single Hidden Layer Neural Networks

研究了一种新的空天飞行器鲁棒自适应轨迹线性化飞行控制系统设计方案。利用单隐层神经网络的逼近能力在线估计系统中存在的不确定性,神经网络输出用以抵消不确定性对轨迹线性化方法控制性能的影响。鲁棒自适应控制器用以克服逼近误差,并使闭环系统具有更好的性能。严格的理论证明表明,给定的自适应调节律能够保证闭环系统跟踪误差最终收敛至任意小紧集。空天飞行器高超声速飞行条件下的仿真结果表明,即使在很恶劣的条件下,新方法仍然表现出很好的响应性能。

A design scheme of an adaptive trajectory linearization control system for an aerospace vehicle was presented by using single hidden layer neural networks（SHLNN）. The existing uncertainties in the system were evaluated on-line by the approximation abilities of SHLNN, the effects of the un- certainties on the system performance were canceled by the neural network output. The approximation error was overcome by a robust adaptive controller to obtain a better perform of the closed-loop control system. A rigorous proof demostrates that the adaptive adjustment rules can ensure converging the system track error to an arbitrarily small neighborhood of zero. The simulation results for the aerospace vehicle show that the on-line neural network based adaptive algorithm has better response performance under the condition of hypersonic flight.