Knowledge Management System Of Institute of optics and electronics, CAS
|Place of Conferral||北京|
|Keyword||机动目标 时滞补偿 预测跟踪 机器学习 Smith预估器|
With the fast development of science and technology, the maneuverability of the target is increasing, and the tracking ability of the electro-optical tracking system can’t meet the current actual requirements and is required to be further improved, especially, for the appearance of new targets such as the small unmanned air vehicle. However, there are many factors that restrict the improvement of tracking ability of the electro-optical tracking system, in which the time delay of image sensor is a major factor. In this paper, the tracking ability of the system is improved by the time delay compensation.
In terms of the control structure, the classical control only includes feed-forward and feedback, and all the control methods are composed by those two basic forms. Therefore, those two control structures are researched to compensate the time delay. By using feed-forward to compensate time delay, it is necessary to predict the velocity or acceleration of the target by estimation means, and then feed it to the velocity control loop or acceleration control loop. By using the feedback to compensate time delay, it is necessary to provide the model of the control plant, which can be used to compensate the delayed feedback signal.
In order to predict the velocity and acceleration of the target accurately, this paper has studied the predictive filtering algorithm. For the tracking of the small unmanned air vehicle, the difficulty is the accuracy of the target motion model. The interacting multi model algorithm has been analyzed and researched for the good prediction accuracy. However, the calculation of IMM is complex, and the model switching efficiency is low, therefore, it is not suitable for the actual application. Based on the analysis of the interacting multiple model algorithm, the predictive feed-forward based on machine learning is proposed, concretely, the motions of target are classified by a support vector machine which is kind of machine learning algorithm, and then, the motion model is switched according to the classification results, therefore, the filtering algorithm run only once in a period of iteration. Simulation results show that the amount of calculation of the proposed approach is far less than the interacting multiple model algorithm, and the accuracy of the prediction is ensured at the same time, in addition, the model switching efficiency is better than the interacting multi model algorithm.
Besides, this paper focuses on the research of Smith predictor control, but the traditional Smith predictor control is sensitive to plant parameter variations. Based on this, this paper puts forward the combination of velocity inner loop. A velocity inner loop is added in the traditional Smith predictor control loop, which results in the improvement of robustness of the controlled plant. The effectiveness of the proposed approach is analyzed in theory, and the experimental verification is designed in the fast steering mirror experiment platform. The experimental results show that the improved Smith predictor control is more stable and effective than the traditional predictor Smith control.
|曹政. 基于时滞补偿的机动目标跟踪技术研究[D]. 北京. 中国科学院大学,2017.|
|Files in This Item:|
|基于时滞补偿的机动目标跟踪技术研究.pd（1850KB）||学位论文||开放获取||CC BY-NC-SA||Application Full Text|
|Recommend this item|
|Export to Endnote|
|Similar articles in Google Scholar|
|Similar articles in Baidu academic|
|Similar articles in Bing Scholar|
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.