保证收敛条件下运载火箭实时精确着陆制导研究

运载火箭可重复使用是人类有效降低进入太空成本的重要途径，是各航天大国重点发展方向。火箭回收在环境/模型不确定性与多约束条件下对着陆精度要求非常高，传统解析制导律或基于离线轨迹跟踪的制导律已无法满足需求，对基于在线复杂非线性优化的制导算法的收敛性、实时性与鲁棒性提出了严峻挑战。提出基于近似解耦与固定部分变量的高维非线动力学降维方法，降低非线性优化的复杂度，在此基础上，提出一种新颖的伪双层优化方法，保证优化算法的收敛性；研究优化模型的改进方法及底层定制高效内点法，提高优化算法的计算效率；提出采用Farkas’引理分析多重不确定性下基于闭环凸优化（含二次约束）的制导算法的鲁棒性，研究实现制导算法鲁棒性的前提条件与解决方法；建立六自由度仿真系统，对研究方法有效性进行验证。研究成果将为我国运载火箭回收的核心精确着陆制导技术提供理论依据，并具有实际工程应用价值。

Reusable launch vehicle development is an important way for human to effectively reduce the cost of access to space, and is a key direction for all nations who significantly value space exploration. In the presence of environmental/model uncertainties and multiple constraints, rocket recycling has high requirements on landing accuracy. This makes traditional analytical guidance laws or those based on offline trajectory-tracking invalidated, and puts forward a severe challenge on convergence, real-time performance, and robustness of guidance algorithms based on onboard nonlinear programming. This project proposes approximated decoupling and partial variable fixing to reduce the dimension of high-dimensional nonlinear systems, which can reduce the complexity of nonlinear programming. Moreover, a new pseudo-bilevel programming method is proposed to guarantee convergence of the optimization algorithm. Besides, this project studies how to ameliorate the optimization model and develop a customized and efficient interior point method, which is to improve the computational efficiency of optimization algorithms. A guidance algorithm can be constructed based on closed-loop convex optimization which contains quadratic constraints, and this project proposes using the Farkas’ lemma to analyze its robustness. Prerequisites for robustness and the corresponding methods for realizing robustness are studied as well. These research methods are then tested by constructing a 6-DOF simulation system. The results in this project will provide theoretical foundation for the key precise landing guidance technologies of rocket recycling in our nation, and are valued to be implemented in practical engineering applications.