基于灰色RBF神经网络模型的电梯层间交通分布预测

为了有效预测电梯的层间交通分布状态,提出一种层间交通O-D矩阵的预测方法．该方法融合灰色预测和神经网络方法各自的优点,将灰色预测方法与RBF神经网络有机结合,构造灰色神经网络预测模型．利用灰色预测中的累加生成运算（accumulatedgeneratingoperation,AGO）对原始观测数据进行变换,得到规律性较强的累加数据,作为神经网络的建模和训练样本．还提出了对不良交通需求数据的修正方法,以进一步降低观测数据的随机性．所提方法既避免了灰色预测方法存在的理论误差,又提高了神经网络的训练速度和预测精度,适用于短期层间交通分布预测．仿真试验验证了该方法的有效性．     

具有敏感区域的华南虎图像配准研究

对数极坐标映射是一种具有刚体不变性的模板匹配算法,以它获得的刚体变换矩阵为初值经LMA(Levenberg Marquardt algorithm)迭代可以实现图像的高精度配准,但对于场景发生剧烈变化的图像对,上述方法无法获得正确的配准参数。为此,提出了一种具有敏感区域的图像配准算法,在LMA迭代中仅考虑落在人工选取的敏感区域中的像素点,通过对陕西林业厅公布的华南虎照片进行的图像配准实验表明了该算法的有效性。     

Application of the general elimination method in robot kinematics

A general method for solving the robot inverse kinematics problem is presented. The method is based upon the general elimination method to obtain the equivalent system of equations which are triangularized and the solutions of the inverse kinematics problem can be solved by backsubstitutions.     

一种基于SPWVD-WVD的高质量时频分析方法及ISAR成像应用

交叉项干扰抑制与高时频聚集度是准确反应信号的时频分布特征的重要因素。传统的魏格纳-维尔分布(Wigner-Ville Distribution,WVD)算法虽能获得较高的时频分辨特征,但分析多成分信号时存在严重的交叉项干扰问题,限制了其实用性;而平滑伪魏格纳-维尔分布(Smoothed Pseudo Wigner-Ville Distribution,SPWVD)算法虽在一定程度上抑制交叉项干扰,但降低了时频聚集度。为了解决上述问题,提出了基于SPWVD-WVD的时频分析方法。该方法利用SPWVD与WVD之间的滤波互消效应,将SPWVD二值化结果与WVD结果进行矩阵运算,最终得到高质量的时频分析结果。实验结果表明,所提出的算法能够有效去除多分量信号的交叉项干扰,提高信号分析结果的时频聚集度,还原多分量信号的真实时频分布。最后将该算法成功应用于逆合成孔径雷达成像中。     

Policy Iteration for Optimal Control of Discrete-Time Time-Varying Nonlinear Systems

Aimed at infinite horizon optimal control problems of discrete time-varying nonlinear systems, in this paper, a new iterative adaptive dynamic programming algorithm, which is the discrete-time time-varying policy iteration (DTTV) algorithm, is developed. The iterative control law is designed to update the iterative value function which approximates the index function of optimal performance. The admissibility of the iterative control law is analyzed. The results show that the iterative value function is non-increasingly convergent to the Bellman-equation optimal solution. To implement the algorithm, neural networks are employed and a new implementation structure is established, which avoids solving the generalized Bellman equation in each iteration. Finally, the optimal control laws for torsional pendulum and inverted pendulum systems are obtained by using the DTTV policy iteration algorithm, where the mass and pendulum bar length are permitted to be time-varying parameters. The effectiveness of the developed method is illustrated by numerical results and comparisons.      

Inverse-compensation-based DETC for nonlinear hysteresis system with disturbance

To control a nonlinear system with both hysteresis and disturbance, it is necessary to establish a hysteresis model and improve the disturbance rejection ability. However, the input signal implicitly involved in the classical hysteresis model can lead to difficulty in constructing a compensator. In this study, a hysteresis model in explicit form is proposed with a bounded auxiliary variable. Then, a model-based inverse is constructed for approximate compensation for the hysteresis. Moreover, the compensation error, which is considered a part of the disturbance, is proved to be bounded. Disturbance estimation triggered control (DETC) is utilized to address the compensation error coupled with the external disturbance. According to the disturbance effect indicator (DEI), DETC can improve the system control performance by considering the disturbance effect judgment. Experimental results are presented to illustrate the potential of the proposed technique.     

特征挖掘与区域增强的弱监督时序动作定位

弱监督时序动作定位旨在定位视频中行为实例的起止边界及识别相应的行为。现有方法尽管取得很大进展,但依然存在动作定位不完整及短动作的漏检问题。为此,提出了特征挖掘与区域增强（FMRE）的定位方法。首先,通过基础分支计算视频片段之间的相似分数,并以此分数聚合上下文信息,得到更具有区别性的段分类分数,实现动作的完整定位;然后,添加增强分支,对基础分支定位中持续时间较短的动作提案沿时间维度进行动态上采样,进而采用多头自注意机制对动作提案间的时间结构显式建模,促进具有时间依赖关系的动作定位且防止短动作的漏检;最后,在两个分支之间构建伪标签互监督,逐步改进在训练过程中生成动作提案的质量。该算法在THUMOS14和ActivityNet1.3数据集上分别取得了70.3%和40.7%的检测性能,证明了所提算法的有效性。     

少样本深度学习算法赋能二维衍射超材料逆向设计

二维衍射超材料是研究高阶衍射光束的光学手性响应的理想材料。随着计算机软硬件技术的发展,越来越多的深度学习算法应用于超材料这一领域。当前,深度学习算法大多需要大量训练数据作为支撑,造成了较大时间开销,并且其模型的可解释性通常较弱。针对以上情况,本文提出一种基于数据增强的迭代少样本深度学习算法,利用少量的仿真数据,训练正向预测的网络模型,进而生成大量“伪数据”,再使用扩充后的混合数据集进行逆向设计模型训练,通过多次迭代得到超材料结构参数的准确值。与以往大多数深度学习方法相比,本文方法将仿真数据集规模缩小了至少一到两个数量级,有效降低了对仿真光谱的依赖,且设计成功率不低于96%。同时,从数据的角度,根据二维衍射超材料的几何参数对光学响应的影响解释了实验结果,增强了模型的可解释性。结果表明,本文方法为超灵敏、可调控偏振光电子器件的研发提供了新思路。     

基于B2B-MMC的柔性互联变电站优化调度方法

针对传统变电站内主变之间采用单母线分段开关难以准确调控潮流问题,提出一种基于背靠背模块化多电平换流器（B2B-MMC）的柔性互联变电站优化调度方法。首先,提出含柔性互联变电站的配电网架构,实现潮流的连续性控制及多区域间互相协调;然后,设计柔性互联变电站架构下的一种潮流优化调度方法,采用遗传算法对换流器进行赋值,利用改进交替迭代法进行潮流计算,实现配电网潮流的主动调控;最后,通过仿真模型验证了本文所提控制方法的有效性和合理性。