宇航学报 ›› 2022, Vol. 43 ›› Issue (4): 423-433.doi: 10.3873/j.issn.1000-1328.2022.04.004

• 飞行器设计与力学 • 上一篇    下一篇

基于马尔可夫蒙特卡洛法的系统辨识方法研究及应用

曹瑞,刘燕斌,陆宇平   

  1. 1. 南京航空航天大学自动化学院,南京 211106;2. 南京航空航天大学航天学院,南京 211106
  • 收稿日期:2021-06-28 修回日期:2021-09-14 出版日期:2022-04-15 发布日期:2022-04-15
  • 基金资助:
    国家自然基金(11572149, 61873126); 中央高校基本科研业务费专项资金(NS2021061); 中国博士后科学基金(2020M681586); 江苏省自然科学基金(BK20200437); 南京航空航天大学博士生跨学科创新基金(KXKCXJJ202008)

Research and Application of System Identification Method Based on Markov Chain Monte Carlo Method

CAO Rui, LIU Yan bin, LU Yu ping   

  1. 1. College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China;2. College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China
  • Received:2021-06-28 Revised:2021-09-14 Online:2022-04-15 Published:2022-04-15

摘要: 提出了基于马尔可夫蒙特卡洛(MCMC)的贝叶斯辨识方法,以解决高超声速飞行器系统辨识中复杂动力学模型转换为简单或稀疏模型所带来的不确定性问题,以及存在的训练数据大和积分难处理的问题。该方法将数据退火算法引入MCMC中,不仅解决了MCMC易陷入局部最优的问题,并且将数据退火与“高信息训练数据”的概念相结合,能够以较低的计算成本分析大数据集。此外,该方法可以对参数估计过程中存在的不确定性进行量化,获得未知参数的最优估计值。通过仿真实验,验证了提出的系统辨识方法的有效性,辨识出的模型能够有效应用于控制器设计之中,并获得较好的控制效果。

关键词: 高超声速飞行器, 系统辨识, 马尔可夫蒙特卡洛(MCMC), 贝叶斯, 不确定性

Abstract: In this paper, a Bayesian identification method based on Markov chain Monte Carlo is proposed to solve the problems of hypersonic vehicle system identification, including the uncertainties caused by transforming complex dynamic model into simple or sparse model, large training data and complex integration. In this method, the data annealing algorithm is introduced into MCMC, which not only solves the problem that MCMC is easy to fall into local optimization, but also combines data annealing with the concept of “high information training data” analyze large data sets with low computational cost. In addition, this method can quantify the uncertainty in the process of parameter estimation and obtain the optimal estimation value of unknown parameters. Through simulation experiments, the effectiveness of this system identification method proposed in this paper is verified, and the identified can be used in controller design and has good control effect.

Key words: Hypersonic vehicle, System identification, Markov chain Monte Carlo (MCMC), Bayesian, Uncertainty

中图分类号: