基于智能自适应技术的RLV再入姿轨协同容错控制

The reentry phase of RLV is the most critical phase for the whole flight process. Due to its complexity of environmental features, the attitude and trajectory coordinate control system is in urgent need and with great challenges. This project applies artificial intelligence algorithm such as brain neural network to the RLV reentry attitude and trajectory coordinate control theory in an innovative way to solve the RLV safe and stable reentry in the presence of dynamic uncertainty and fault influence. Through the establishment of model parameter database, an on-line parameter rapid identification method based on a kind of brain-like neural network is put forward to obtain RLV precise reentry model; Considering the practical problems during the RLV reentry phase, based on the event-driven principle, an adaptive Gauss pseudo-spectral method combined with improved particle swarm optimization algorithm is proposed to solve RLV trajectory online autonomous optimization and tracking problem in the dynamic uncertain environment; On this foundation, an integrated attitude controller and fault observer design method for reentry RLV is proposed based on adaptive dynamic programming and a new set of trajectory and attitude intelligent collaborative fault-tolerant control system is also build; Through the introduction of intelligent network evaluation, online autonomous optimization and attitude & trajectory coordinate correction strategy, RLV can still achieve smooth reentry flight intelligently and autonomously even in the face of strong interference, the actuator fault and complex dynamic uncertain environment, which is supposed to be forward-looking work in theory.

RLV再入过程是整个飞行过程中最为关键的阶段，由于环境特性复杂，使得智能、自主、可靠的姿轨控制系统需求迫切且面临巨大挑战。本课题创新性的将类脑神经网络等智能算法应用于RLV再入姿轨协同控制当中，解决动态不确定和故障影响下的RLV安全稳定再入问题。通过建立模型参数的数据库，提出一种基于类脑神经网络的在线参数快速辨识方法，从而获得RLV精准的再入模型；考虑RLV再入过程面临的工程实际问题，基于事件驱动原则，将自适应高斯伪谱法和改进粒子群算法进行结合，解决动态不确定环境下的RLV轨迹在线自主优化与跟踪问题；在此基础上，提出基于自适应动态规划的RLV再入姿态控制器和故障观测器综合设计方法，构建一套新的RLV轨迹与姿态智能协同容错控制系统，通过引入智能网络评价、在线自主优化、姿轨协调修正等策略，使得飞行器在面临强干扰、执行器故障和复杂不确定动态环境时，仍能智能自主地实现平稳再入飞行，在理论上具有一定的前瞻性。

针对航天运载系统“快速、机动、可靠、低成本”的需求，可重复使用运载器（RLV，Reusable Launch Vehicle）已成为目前世界各航天大国正在部署研制的可天地往返、可重复使用的新型空天飞行器。为实现RLV安全、稳定、自主再入飞行，本项目对RLV再入建模、轨迹优化与跟踪、姿轨协同容错控制三个核心问题进行了深入研究。首先，构建了一套RLV再入的模型参数知识库，实现了基于知识库的参数在线学习和辨识方法，获得了较为精确的RLV再入模型，为在线实时轨迹优化与姿轨协同控制提供了必要的模型数据支撑。其次，根据安全事件驱动原则，结合RLV复杂轨迹优化模型，提出了求解速度快、解算精度高的实时优化方法和策略，解决了动态不确定环境下的轨迹自主在线重规划和精确跟踪问题。最后，综合考虑轨迹优化与姿态跟踪控制的复杂耦合关系，提出了一种基于故障观测器的智能自适应姿轨协同控制方法，构建了一套轨迹与姿态协同容错控制系统，实现了轨迹精确跟踪与姿态快速稳定。本项目建立了RLV智能控制优化理论与应用方法，通过人工智能技术在RLV再入姿轨协同控制方面的创新应用，实现了以工程应用为背景的理论基础研究与创新，解决了RLV再入自主性、可靠性和灵活性的迫切需求，提高了RLV的智能自主飞行能力。