Vehicle active steering control research based on two-DOF robust internal model control

Because of vehicle’s external disturbances and model uncertainties, robust control algorithms have obtained popularity in vehicle stability control. The robust control usually gives up performance in order to guarantee the robustness of the control algorithm, therefore an improved robust internal model control(IMC) algorithm blending model tracking and internal model control is put forward for active steering system in order to reach high performance of yaw rate tracking with certain robustness. The proposed algorithm inherits the good model tracking ability of the IMC control and guarantees robustness to model uncertainties. In order to separate the design process of model tracking from the robustness design process, the improved 2 degree of freedom(DOF) robust internal model controller structure is given from the standard Youla parameterization. Simulations of double lane change maneuver and those of crosswind disturbances are conducted for evaluating the robust control algorithm, on the basis of a nonlinear vehicle simulation model with a magic tyre model. Results show that the established 2-DOF robust IMC method has better model tracking ability and a guaranteed level of robustness and robust performance, which can enhance the vehicle stability and handling, regardless of variations of the vehicle model parameters and the external crosswind interferences. Contradiction between performance and robustness of active steering control algorithm is solved and higher control performance with certain robustness to model uncertainties is obtained.     

AQM controller design for networks supporting TCP vegas: a control theoretical approach

In this paper, a mathematical model and control theoretical framework for designing AQM controllers in networks supporting TCP Vegas is introduced. We have emphasized on a modified TCP Vegas algorithm that can respond to congestion signals through explicit congestion notification (ECN). The overall nonlinear delayed differential equations of the dynamics model of closed loop system have been derived based on TCP Vegas model. The model is then linearized to derive a transfer function representation between the packet marking probability and the bottleneck router queue length as the input and output of the modified TCP Vegas/AQM system. The model properties have been then examined especially for the case of single bottleneck homogeneous network which is closely investigated. Finally an AQM controller based on Coefficient Diagram Method (CDM) has been designed for the system and its performance has been compared with some other AQM controllers. CDM is a new indirect pole placement method that considers the speed, stability and robustness of the closed loop system in terms of time domain specifications. In order for synthesizing the simulation scenarios, our campus router traffic has been studied experimentally for a sample period of one hour and the corresponding parameters has been extracted. The simulation results are representative of good performance of developed TCP Vegas/AQM structure for different simulated scenarios.     

基于深度强化学习多用户移动边缘计算轻量任务卸载优化

移动边缘计算(MEC)在提高移动设备的计算体验质量方面具有一定的应用前景.它可以为支持传统通信和MEC服务的切片式无线接入网提供紧密邻近的计算功能.然而,这种密集计算问题是一种高维的NP难问题,一些机器学习方法在解决该问题的时候不能取得良好的效果.针对这些问题,本文将最佳计算卸载问题建模为马尔可夫决策过程,目标是最大化长期效用性能,从而根据队列状态、能量队列状态以及移动用户与BS之间的信道质量做出卸载决策.为了降低状态空间中高维性的问题,提出了应用深度确定性策略梯度的基于候选网络优化边缘计算优化卸载ECOO算法,从而产生一种用于解决随机任务卸载的新型学习算法.通过仿真实验证明,ECOO算法在能耗和时延方面优于一些深度强化学习算法,在处理高维问题时效果更好.     

数控机床热误差时间序列模型预测稳健性的提升

针对数控机床热误差建模应用的时间序列算法受严重多重共线性的影响存在预测稳健性不足的问题,提出一种提升时间序列预测稳健性的方法。该方法将时间序列算法与能够抑制多重共线性的建模算法相结合,从而既可通过在模型中加入温度滞后值来提供更全面的温度信息,又可对温度滞后值引入的更为严重的多重共线性进行处理。文中以时间序列算法中的分布滞后(DL)算法、共线性抑制算法中的主成分回归(PCR)算法为例,采用主成分分布滞后(PCDL)算法建立了机床热误差补偿模型,并将其与DL算法的预测精度和稳健性进行了比较。结果显示,PCDL算法因为抑制了多重共线性的影响,其模型预测精度和稳健性远优于DL模型,预测精度提升了约9μm。本文所述方法可为时间序列数据建模在不同领域内的应用提供参考。     

携带队列长度信息的ABR业务流量控制方案

本文提出了一种新的基于EFCI的ABR流量控制算法--E-EFCI,它的基本思想是扩展EFCI的功能,使其不仅反映网络阻塞状况.而且还反映网络排队状况。在交换机中通过队列的长度设置EFCI来传递网络的排队信息。仿真结果表明.本文提出的算法比基本EFCI算法具有更好的控制行为,减小了ACR和排队长度的振荡幅度、提高了链路、缓存利用率,加快拥塞解除,并具有较高的性能价格比。     

Effects of defocus and algorithm on optical step height calibration

Defocus effects on step height measurements by interferometric microscopy are estimated using different algorithms to calculate the step height. The interferometric microscope here is a Mirau-type with a 20x objective and a numerical aperture (NA) of 0.4. Although the focus is adjusted within the range of the depth of focus, a defocus corresponding to 4 fringes (1.3 μm) distorts the measured profile into a curve with a radius of curvature of about 24 m. The effect of this distortion on step height determination by a basic ISO step height algorithm is estimated to be 0.11 nm/fringe. An algorithm developed by NIST and a second-order fitted ISO step-height algorithm show good robustness against defocus errors. Because of sample imperfection, a discrepancy of about 1 % exists among step-height values determined by the different algorithms at the optimum fringe contrast position. It has been pointed out theoretically that the NA correction factor varies with the difference of optical path length between the sample and the reference surface. However, in our case, the changes of optical path length do not change the NA correction factor by more than 0.3 %.     

基于深度学习的铸件CT图像分割算法

针对现有方法分割弱边缘铸件CT图像难度大、精度低、鲁棒性差的问题,提出一种融合残差模块与混合注意力机制的U型网络分割算法(AttRes-U-Nets)。该算法以U-Net网络为基础,首先构建深度残差网络ResNets作为算法的编码网络,解决传统U-Net网络特征提取能力不足的问题;然后,引入改进后的混合注意力机制,突出分割目标区域与通道的特征响应,提高网络灵敏度;最后,将Focal loss与Dice loss结合为一种新损失函数FD loss缓解样本不平衡带来的负面影响。使用120阀体数据集对算法性能进行验证,实验结果表明,本文算法对铸件分割的像素准确率(PA)和交互比(IoU)分别达到98.72%和97.40%,优于传统U-Net算法与其他主流语义分割算法,为弱边缘分割提供了新思路。     

改进的协同粒子群优化算法的研究与应用

为增强现有PSO算法和协同粒子群优化算法的优化性能,提出了一种改进的协同粒子群优化算法及一种新的协同策略。该算法在进化过程中,将寻优粒子群分解为若干子分群,各子分群粒子利用本分群经验和整个种群经验进行搜索,既能在分群内部不断搜索,不迷失寻优方向,又能周期性地共享整群最优值引导粒子找到最好解。分解为多个子种群有利于维持种群的多样性,有效抑制局部最优现象发生。对经典复杂函数的寻优测试表明,改进算法的鲁棒性、收敛速度、精度及全局搜索能力均优于基本PSO算法。最后将改进算法用于建立基于神经网络的旋转机械故障诊断模型,设计了相应的故障诊断系统。结果表明,基于此算法的故障诊断系统具有诊断精度较高、稳定性能较好等特点。     

基于遗传神经网络的加速度传感器动态建模方法

提出了利用遗传神经网络实现加速度传感器动态建模的新方法，介绍动态建模原理以及算法，给出用遗传神经网络建立的加速度传感器动态数学模型。该方法利用加速度传感器的动态标定数据，采用遗传神经网络搜索和优化动态模型参数。这样，既保留遗传算法的全局搜索能力，克服神经网络容易陷入局部极小的缺陷，又具有神经网络局部搜索能力强的特点。结果表明：以上提出的动态建模方法具有建模精度高、鲁棒性好等优点。     

Locating service centers optimizing customers’ perspective criteria

A network of several demand nodes is considered. We want to locate a number of multiserver facilities at some nodes assuming at least one server must be located at each node. Accordingly, the number of servers has to be more than the nodes number. We propose a model for locating these facilities from customers’ perspective in such a way as to minimize the total average queue length, when customers choose the facilities they patronize, by the travel distance (or time) to the facility and the number of servers at the facility. This means that each customer can obtain the service or goods from several (i.e., rather than only one) facilities, according to a probabilistic distribution. Genetic and simulated annealing (SA) algorithms based heuristics are proposed to solve this problem, which belongs to nonlinear integer programming problems. We evaluate the proposed algorithms for the solution quality solving a number of numerical problems. The results indicate that SA whose objective function is around 26% less than that of genetic is of higher performance.     

基于模糊神经网络PID的液位串级控制的研究

利用PID算法对液位串级系统进行控制虽然是一种有效的控制方法,但由于它的精确数学模型难以确定,使得参数整定困难、控制效果不理想。该文将PID算法、模糊控制算法以及神经网络算法相结合,形成了一种智能控制算法——模糊神经网络PID算法。将该算法运用到液位串级控制系统中,实现了PID参数的自整定,并且提高了控制质量。实验结果表明,模糊神经网络PID算法与PID算法的控制效果相比,在鲁棒性和响应时间等方面有了较大的提高,具有一定的应用前景。     

Predictive functional control for active queue management in congested TCP/IP networks

Predictive functional control (PFC) as a new active queue management (AQM) method in dynamic TCP networks supporting explicit congestion notification (ECN) is proposed. The ability of the controller in handling system delay along with its simplicity and low computational load makes PFC a privileged AQM method in the high speed networks. Besides, considering the disturbance term (which represents model/process mismatches, external disturbances, and existing noise) in the control formulation adds some level of robustness into the PFC-AQM controller. This is an important and desired property in the control of dynamically-varying computer networks. In this paper, the controller is designed based on a small signal linearized fluid-flow model of the TCP/AQM networks. Then, closed-loop transfer function representation of the system is derived to analyze the robustness with respect to the network and controller parameters. The analytical as well as the packet-level ns-2 simulation results show the out-performance of the developed controller for both queue regulation and resource utilization. Fast response, low queue fluctuations (and consequently low delay jitter), high link utilization, good disturbance rejection, scalability, and low packet marking probability are other features of the developed method with respect to other well-known AQM methods such as RED, PI, and REM which are also simulated for comparison.     

An optimal general type-2 fuzzy controller for Urban Traffic Network

Urban traffic network model is illustrated by state-charts and object-diagram. However, they have limitations to show the behavioral perspective of the Traffic Information flow. Consequently, a state space model is used to calculate the half-value waiting time of vehicles. In this study, a combination of the general type-2 fuzzy logic sets and the Modified Backtracking Search Algorithm (MBSA) techniques are used in order to control the traffic signal scheduling and phase succession so as to guarantee a smooth flow of traffic with the least wait times and average queue length. The parameters of input and output membership functions are optimized simultaneously by the novel heuristic algorithm MBSA. A comparison is made between the achieved results with those of optimal and conventional type-1 fuzzy logic controllers.     

基于增强卷积神经网络的电子鼻长期漂移抑制方法

为解决电子鼻传感器阵列中的漂移问题,提出了一种增强卷积神经网络的长期漂移抑制方法.首先,通过结合历史数据的方式进行数据库扩增,起到了数据增强的效果;然后,使用增量补偿模块结合增量学习思维进行网络训练,起到了模型增强的效果;最后,分别使用公开数据集和实测数据集来验证模型的漂移抑制效果.实验结果表明:增强卷积神经网络算法的...     

基于CMAC神经网络PID控制算法控制直流电机的仿真研究

传统的PID控制算法对控制参数难以适应,抗干扰能力差,对直流电动机进行控制时速度较慢、稳定性较差,为解决上述问题,文中提出了一种基于CMAC神经网络的PID控制算法来控制直流电动机,对PID控制参数进行自适应修改,仿真结果表明,该算法提高了系统的稳定性、响应速度、参数适应性和鲁棒性,改善了系统控制存在的稳态精度不高的问题。     

堆垛机速度智能控制的优化研究

采用模糊控制技术与小脑模型神经网络(CMAC)相结合的方式进行堆垛机的速度控制,克服单独运用模糊控制或CMAC神经网络的缺点,使系统既具有模糊控制的灵活性和强适应性,又兼具神经网络的学习能力,并且采用遗传算法对控制器的输入输出比例因子及连接权值进行寻优.仿真结果表明:该控制系统提高了系统的稳定性、鲁棒性和控制精度,使系统的综合性能得到显著改善.     

一种新的动态神经网络及其在机器人监督控制中的应用

针对提高机器人控制性能这一类问题，在复合输入动态递归网络的基础上，提出一种新的动态递归神经网络结构，称为状态延迟输入动态递归神经网络（State Delay Input Dynamical Recurrent Neural Networks)。该动态网络具有新的拓扑结构及学习规则，各权值矩阵的含义更为明确，权值的训练过程更为简洁。网络增加了输入输出层前一步的状态信息，收敛速度及稳态精度与其它常用网络结构相比均有明显提高。将该网络用于机器人的监督控制系统，利用神经网络建立起被控对象的逆模型，与传统PD控制器结合，进一步确保了控制系统的稳定性，有效地提高系统的精度和自适应能力。仿真结果表明了这种改进的有效性和优越性。     

遗传参数自适应调整算法及其在液压动力系统优化控制中的应用

针对固定参数的遗传算法容易陷入过早收敛,进入局部最优状态等问题,建立了交叉概率及变异概率的模糊逻辑控制器以实现遗传算法策略性参数的自适应调整,从而提高优化算法的收敛速度及获得全局解的能力。运用常规优化方法及改进优化算法对永磁电机驱动的液压系统流量进行优化控制和对比,仿真和实验结果表明：采用遗传参数自适应调整算法优化控制器,可使系统在典型工况下,保持良好的控制性能,并且具有高于常规优化方法的控制精度和鲁棒性。     

H.264/AVC压缩域鲁棒视频水印

针对现有H.264/AVC压缩域开环水印方法存在误差漂移及鲁棒性差的问题,分析了该类水印算法的特点及误差漂移机制,提出了一种无误差漂移的鲁棒视频水印算法。首先,基于视频重建过程,分析了开环回路水印方法的误差漂移机制,得到由水印信息造成的独立重建误差。然后,采用误差间的线性组合得到4个可以避免误差漂移的水印模板。最后,通过调制残差系数与相应水印模板之间的正负相关性,将水印嵌入在非零系数多的4×4子块上,而不像以往算法将水印直接添加在残差系数上。实验结果表明:所提算法造成的平均结构相似度下降在0.005以内;码率的增长在1.00%左右;对重量化转码、加性高斯白噪声、亮度调节等常见水印攻击,误码率均在0.15以下。得到的结果满足视频水印算法对透明性,码率稳定性,鲁棒性的要求。