无尾飞翼布局飞行器的操纵面故障强化学习最优自适应补偿控制研究

项目名称： 无尾飞翼布局飞行器的操纵面故障强化学习最优自适应补偿控制研究

项目编号： No.61473147

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 张绍杰

作者单位： 南京航空航天大学

项目金额： 83万元

中文摘要： 本项目在对无尾飞翼布局飞行器操纵面故障典型飞行特性分析的基础上，挖掘其蕴涵的科学问题，研究MIMO仿射非线性系统的执行器故障强化学习最优自适应补偿控制方法，改善容错控制系统的性能。首先，采用CATIA建模、CFD和AMS等分析方法研究无尾飞翼布局飞行器操纵面故障引起的空气动力不平衡和失速等飞行特性，总结主要容错控制问题。其次，针对具有执行器故障的MIMO仿射非线性系统，提出基于MMST的执行器分组方法和系统PPB动态性能分析方法。然后，将非线性系统的actor-critic结构强化学习自适应最优控制用于自适应补偿控制，提出了最优自适应补偿控制方案；针对存在执行器饱和的情况，基于扇形区域分析和双线性矩阵不等式方法设计抗饱和补偿器，实现系统的最优自适应补偿控制。最后，利用AMESim、Matlab和Visual C++等工具开发了无尾飞翼布局飞行器容错飞行控制仿真平台。

中文关键词： 无尾飞翼布局飞行器；非线性系统；最优自适应补偿控制；强化学习；执行器故障

英文摘要： Based on the scientific issues analyzing of typical fault tolerant flight control (FTFC) problems for tailess flying wing aircrafts with control surface failures, some control schemes which can improve the control performance of the fault tolerant control systems are proposed by reinforcement learning optimal adaptive compensation control method for multi-input multi-ouput (MIMO) affine nonlinear systems with actuator failures in this project. Firstly, the air dynamic imbalance and stall characteristics of tailess flying wing aircrafts with control surface failures are analysed by means of CATIA, CFD and AMS. Secondly, more effective actuator grouping way is proposed by means of multiple models switching and tuning (MMST) method, representation of the system dynamic performance is studied by prescribed performance bounds (PPB) analysis. Thirdly, a new control scheme named optimal adaptive compensation cntrol are proposed by using of actor-critic reinforcement learning adaptive optimal control and adaptive compensation control methods. Then, an optimal adatpive compensation control scheme is proposed for the system with actuator satuaration by anti-windup compensators. The anti-windup compensators of the controlled systems are designed with sector bound condition and bilinear matrix inequality analysis. Finally,the theoretical results are used to a tailess flying wing aircraft flight control system, and a fault tolerant flight control simulation platform is designed by AMESim, Visual C++ and Matlab.

英文关键词： tailess flying wing aircrafts;nonlinear systems;optimal adaptive compensation control;reinforcement learning;actuator failures