基于混沌神经网络的QoS组播路由研究

发布时间：2018-10-26 17:42

【摘要】：组播是指一个信息源点传输到多个目标节点的的信息传输方式,QoS(Quality of Sevice)称为服务质量,是一种网络安全机制,用来解决网络延迟和阻塞等问题,是指网络提供更高优先服务的一种能力。随着新型网络业务大量涌现,带服务质量保证的组播技术成为研究热点。QoS组播路由问题又称Steiner树问题,用来使组播树成本最小化,已被证明是NP完全问题。选择合适的QoS组播路由算法对于高质量的组播通讯具有重要意义,混沌神经网络算法便是求解此类问题的一种有效方法。以往的混沌神经网络求解QoS组播路由问题多侧重于改进神经网络结构提升算法性能,而忽略了对能量函数的改进,无法对输出矩阵的“行”“列”项进行严格约束。本文在传统能量函数的基础上添加了两个新的约束项,构造出了新的能量函数,保证了闭合路径的有效性。将改进能量函数与暂态混沌神经网络相结合求解QoS组播路由问题。仿真结果表明,改进的算法能够有效提高网络收敛到最优解的概率和速度,且同时适用于复杂程度不同的组播网络。噪声混沌神经网络是在暂态混沌神经网络的基础上添加指数衰减的噪声项得到的,具有随机模拟退火特性。本文将改进的能量函数与噪声混沌神经网络相结合求解QoS组播路由问题。仿真结果显示,噪声混沌神经网络可以使有效解率和最优解率上升,但对于不同原因引起的优化效果不佳,随机噪声的改善作用也有所不同。同时,初始噪声幅值与噪声模拟退火速度必须控制在适当的范围内,否则会引起优化效果下降。迟滞噪声混沌神经网络既能够表现出随机混沌模拟退火又能表现出迟滞动力,迟滞动力有助于神经网络跳出局部极值,而在此基础上得到的基于噪声调节因子的迟滞噪声混沌神经网络可实现对随机噪声水平的控制。本文将迟滞噪声混沌神经网络、基于噪声调节因子的迟滞噪声混沌神经网络和改进的能量函数应用于QoS组播路由问题。仿真结果表明,高噪声条件下,逆时迟滞噪声混沌神经网络的优化结果优于噪声混沌神经网络,而在低噪声条件下,应采用顺时迟滞噪声混沌神经网络改善优化结果;基于噪声调节因子的迟滞噪声混沌神经网络拥有更强的迟滞动态,无论噪声水平高低,都能通过控制噪声调节因子获得优于迟滞噪声混沌神经网络和噪声混沌神经网络的优化效果。
[Abstract]:Multicast refers to the information transmission mode (, QoS (Quality of Sevice) from one information source point to multiple target nodes called quality of Service (QoS). It is a network security mechanism used to solve the problems of network delay and congestion. It refers to the ability of the network to provide higher priority services. With the emergence of new network services, multicast technology with quality of service (QoS) assurance has become a research hotspot. QoS multicast routing problem, also known as Steiner tree problem, has been proved to be a complete NP problem to minimize the cost of multicast tree. It is very important to select suitable QoS multicast routing algorithm for high quality multicast communication. Chaotic neural network is an effective method to solve this kind of problem. In the past, chaotic neural networks used to solve QoS multicast routing problems focused on improving the performance of the neural network structure, but neglected the improvement of the energy function, and could not strictly constrain the "row" column of the output matrix. In this paper, two new constraints are added to the traditional energy function, and a new energy function is constructed to ensure the validity of the closed path. The improved energy function and the transient chaotic neural network are combined to solve the QoS multicast routing problem. Simulation results show that the improved algorithm can effectively improve the probability and speed of convergence to the optimal solution, and it is also suitable for multicast networks with different complexity. Noise chaotic neural network is obtained by adding exponentially attenuated noise term on the basis of transient chaotic neural network. It has the property of stochastic simulated annealing. In this paper, the improved energy function and the noisy chaotic neural network are combined to solve the QoS multicast routing problem. The simulation results show that the noise chaotic neural network can increase the efficient and optimal solution rates, but the improvement effect of random noise is different for different reasons. At the same time, the initial noise amplitude and the simulated annealing speed of noise must be controlled within a proper range, otherwise the optimization effect will be reduced. The hysteresis noise chaotic neural network can show both stochastic chaotic simulated annealing and hysteresis dynamics, which can help the neural network to jump out of the local extremum. The chaotic neural network based on noise regulation factor can control the random noise level. In this paper, the hysteretic noise chaotic neural network, the hysteretic noise chaotic neural network based on noise regulation factor and the improved energy function are applied to the QoS multicast routing problem. The simulation results show that the optimization results of chaotic neural networks with inverse hysteretic noise are better than those with noisy chaotic neural networks under high noise conditions, but under low noise conditions, the chaotic neural networks with time-delay noise should be used to improve the optimization results. The hysteretic noise chaotic neural network based on noise regulation factor has stronger hysteresis dynamics, regardless of the level of noise. By controlling the noise regulation factor, the optimization results are better than those of hysteretic noise chaotic neural network and noise chaotic neural network.
【学位授予单位】：齐齐哈尔大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP183;TP393.03

【相似文献】