基于时滞补偿的机动目标跟踪技术研究

发布时间：2018-04-20 09:41

本文选题：机动目标 + 时滞补偿　；参考：《中国科学院大学(中国科学院光电技术研究所)》2017年硕士论文

【摘要】：伴随着科学技术的飞速发展,目标的机动性也在不断地增强,尤其是小型无人机这类新型目标的出现,使得光电跟踪系统的跟踪能力已经很难满足目前的实际需求,需要进一步提高。而限制光电跟踪系统跟踪能力提升的因素有很多,其中图像传感器的时间滞后是一个主要因素,本文从时滞补偿的角度出发来提升系统的跟踪能力。从控制结构来讲,经典控制包含的只有前馈和反馈,所有控制方法都是由这两种基本形式组成。本文从这两个方面出发对滞后进行补偿。利用前馈补偿滞后,需要通过状态估计方法来预测目标的速度或加速度,然后前馈到速度控制回路或加速度控制回路。利用反馈补偿滞后,则需要建立控制对象的预估模型,利用预估模型补偿掉含有滞后的反馈量。为了能准确预测目标的速度或加速度,本文研究了预测滤波算法,对于小型无人机这类目标的跟踪,其难点在于目标运动模型的准确度。分析和研究了当前主流的交互式多模型算法,预测精度有较好的效果,但计算复杂,模型切换效率低,不适合实际使用。基于对交互式多模型算法的分析,提出了基于机器学习的预测前馈,通过机器学习的支持向量机对目标的运动状态进行分类,根据分类结果切换运动模型,滤波算法只需进行一次运算。仿真实验结果表明,提出的基于机器学习的预测前馈算法计算量远远小于交互式多模型算法,同时,保证了预测的精度,在模型切换效率上也优于交互式多模型算法。此外,对于预估补偿本文重点研究了Smith预估控制,但传统的Smith预估控制对控制对象的参数变化很敏感,基于此,本文提出了与速度内回路相结合的方式,在传统Smith预估控制回路中增加一个内回路,相当于改善了控制对象的鲁棒性,并且从理论上分析了改进方案的有效性,在快反镜实验平台上设计了跟踪实验方案,实验结果表明,改进的Smith预估控制比传统的Smith预估控制更加稳定和有效。
[Abstract]:With the rapid development of science and technology, the maneuverability of targets is also increasing, especially the emergence of new targets such as small UAVs, which makes the tracking ability of photoelectric tracking system difficult to meet the actual needs. Further improvements are needed. There are many factors that limit the tracking ability of photoelectric tracking system, among which the time lag of image sensor is the main factor. In this paper, the tracking ability of the system is improved from the angle of time-delay compensation. From the control structure, the classical control contains only feedforward and feedback, all control methods are composed of these two basic forms. This article starts from these two aspects to carry on the compensation to the lag. Using feedforward compensation for hysteresis, it is necessary to predict the velocity or acceleration of the target by state estimation, and then feed forward to the speed control loop or acceleration control loop. Using feedback to compensate the lag, the predictive model of the control object is established, and the feedback with the lag is compensated by the predictive model. In order to accurately predict the velocity or acceleration of the target, the predictive filtering algorithm is studied in this paper. For the tracking of small UAV, the difficulty lies in the accuracy of the target moving model. This paper analyzes and studies the current mainstream interactive multi-model algorithms. The prediction accuracy is good, but the calculation is complicated and the model switching efficiency is low, so it is not suitable for practical use. Based on the analysis of the interactive multi-model algorithm, the predictive feedforward based on machine learning is proposed. The moving state of the target is classified by the support vector machine of machine learning, and the motion model is switched according to the classification results. The filtering algorithm needs only one operation. The simulation results show that the computational complexity of the proposed predictive feedforward algorithm based on machine learning is much less than that of the interactive multi-model algorithm. At the same time, the prediction accuracy is guaranteed and the model switching efficiency is better than that of the interactive multi-model algorithm. In addition, the paper focuses on the Smith predictive control for predictive compensation, but the traditional Smith predictive control is very sensitive to the change of the parameters of the control object. Based on this, this paper proposes a way to combine with the velocity inner loop. Adding an internal loop to the traditional Smith predictive control loop is equivalent to improving the robustness of the control object. The effectiveness of the improved scheme is analyzed theoretically, and the tracking experimental scheme is designed on the fast mirror experimental platform. Experimental results show that the improved Smith predictive control is more stable and effective than the traditional Smith predictive control.
【学位授予单位】：中国科学院大学(中国科学院光电技术研究所)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP181;TN29

【相似文献】