混沌时滞神经网络的同步研究

发布时间：2018-05-23 17:54

本文选题：混沌神经网络 + 变时滞　；参考：《西南大学》2017年硕士论文

【摘要】：如今是一个网络信息发达的时代,资讯随处可见,信息到处可达,然而却隐藏着信息安全的问题。尤其是在公安、军事和国防等相关领域中的保密通信,对国家安全、利益以及人民的生命财产安全有着直接的关系。同时,互联网上的信息传输和日常的电话联络的安全保密性能也受到广大研究者和用户的关注。如何确保信息在传输过程中的安全性,最重要的途径即是在利用相关加密技术对信息进行加密传输。那么,设计一个保密性能强、安全系数高的加密信号至关重要。混沌信号类似噪声一般,具有隐蔽性强、运动轨迹复杂、难以破解和预测等特点,其非常适合在保密通信中充当加密信号的角色。随着神经网络和混沌学的不断发展以及相互渗透,学者们发现,神经网络在某种条件下可以产生混沌现象。混沌神经网络不仅结构简单易于用硬件电路实现,而且具有复杂的混沌动力学行为,能够产生高度复杂且具有无穷维的混沌信号,可以满足保密通信中对加密信号有较高的要求。所以,具有混沌现象的神经网络非常适合做加密信号。其次,神经网络中神经元之间的信息传输通常都会产生时滞。时滞会诱发神经网络产生更加复杂的混沌时间序列,使得加密信息能力的安全系数更高。此外,在接受端要把加密之前的信息提取出来,就需要用到混沌同步技术。由于时滞的引入,加大了同步控制器设计难度。在实际工程中,系统参数不确定也会给同步控制性能带来很大的影响。本文的主要研究工作如下:第一,针对具有参数不确定性的变时滞混沌神经网络,设计鲁棒控制器以保证两个结构相同、初始值不同的混沌神经网络同步。其中,参数不确定性是时变且范数有界。利用采样控制技术,考虑了两个随机发生且概率已知的采样周期。建立了含有随机变量的同步误差状态方程,并构造新的李雅谱诺夫泛函,推导出鲁棒同步的充分条件。通过MATLAB的LIM工具箱,求解得到合适的控制器反馈增益矩阵以保证两个相同结构的具有参数不确定性的混沌时滞神经网络的全局均方鲁棒同步。此外,相比于周期采样,利用随机采样可以得到更大的采样周期。第二,针对同时具有离散时滞和分布时滞的混沌神经网络,同样使用随机采样的控制和输入延迟的方法,设计同步控制器。在基于两个采样周期的基础上,推广到多个采样周期,把两个系统的同步问题转化为含有随机变量的同步误差状态方程的稳定性问题,并且重新构建新的李雅谱诺夫泛函,利用不等式技术和自由权矩阵的方法,得到全局均方同步的充分条件。所得到的结果,比同等模型的周期采样更有优越性。
[Abstract]:Now is a network information developed era, information can be seen everywhere, information can be reached everywhere, but hidden the problem of information security. Especially in the public security, military, defense and other related areas of confidential communications, has a direct relationship with national security, interests and the safety of people's lives and property. At the same time, the information transmission on the Internet and the security performance of daily telephone contact are also concerned by researchers and users. The most important way to ensure the security of information transmission is to use the related encryption technology to encrypt the information. So, it is very important to design an encrypted signal with strong security and high safety factor. Chaotic signal is similar to noise and has the characteristics of strong concealment complex motion trajectory difficult to decipher and prediction etc. It is very suitable to play the role of encrypted signal in secure communication. With the development and mutual penetration of neural network and chaos, scholars find that the neural network can produce chaos under certain conditions. Chaotic neural network not only has simple structure and easy to be realized by hardware circuit, but also has complex chaotic dynamic behavior, which can produce highly complex and infinite dimensional chaotic signals. It can satisfy the requirement of encrypted signal in secure communication. Therefore, the neural network with chaotic phenomena is very suitable for making encrypted signals. Secondly, information transmission between neurons in neural networks usually produces time delay. The delay will induce the neural network to produce more complicated chaotic time series, which makes the security factor of encryption information higher. In addition, chaotic synchronization is needed to extract the information before encryption at the receiving end. Because of the introduction of time delay, it is more difficult to design synchronous controller. In practical engineering, the uncertainty of system parameters will also have a great impact on the performance of synchronous control. The main work of this paper is as follows: firstly, a robust controller is designed for chaotic neural networks with variable delay with parameter uncertainty to ensure synchronization of two chaotic neural networks with the same structure and different initial values. The parameter uncertainty is time-varying and norm bounded. Two random sampling periods with known probability are considered by sampling control technique. The state equation of synchronization error with random variables is established and a new Lyapunov Functionals are constructed. The sufficient conditions for robust synchronization are derived. By using MATLAB's LIM toolbox, an appropriate controller feedback gain matrix is obtained to ensure the global mean square robust synchronization of two chaotic time-delay neural networks with the same structure. In addition, compared with periodic sampling, the random sampling can be used to obtain a larger sampling period. Secondly, for chaotic neural networks with both discrete and distributed delays, a synchronization controller is designed using the method of random sampling control and input delay. On the basis of two sampling periods, the synchronization problem of two systems is transformed into the stability of synchronization error equation of state with random variables, and a new Lyapunov functional is constructed. A sufficient condition for global mean square synchronization is obtained by using inequality technique and free matrix method. The results obtained are superior to the periodic sampling of the same model.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O415.5;TP183

【参考文献】