The distributionally robust complementarity problem

We investigate the linear complementarity problem with uncertain parameters (ULCP) which affect the linear mapping affinely or quadratically. Assuming that the distribution of the uncertain parameters belongs to some ambiguity set with prescribed partial information, we formulate the ULCP as a distributionally robust optimization reformulation named as the distributionally robust complementarity problem (DRCP), which minimizes the worst case of an expected complementarity measure with a joint chance constraint that the probability of the linear mapping being nonnegative is not less than a given level. Applying the cone dual theory and S-procedure, we conservatively approximate the DRCP as a nonlinear semidefinite programming (NSDP) with bilinear matrix inequalities, which can be solved by the NSDP solver PENLAB. The preliminary numerical test on a constrained stochastic linear quadratic control problem shows that the DRCP as well as the corresponding solution method is promising.     

Control design for discrete‐time bilinear systems using the scalarized Schur complement

In this paper, controller design for discrete‐time bilinear systems is investigated by using sum of squares programming methods and quadratic Lyapunov functions. The class of rational polynomial controllers is considered, and necessary conditions on the degree of controller polynomials for quadratic stability are derived. Next, a scalarized version of the Schur complement is proposed. For controller design, the Lyapunov difference inequality is converted to a sum of squares problem, and an optimization problem is proposed to design a controller, which maximizes the region of quadratic stability of the bilinear system. Input constraints can also be accounted for. Copyright © 2017 John Wiley & Sons, Ltd.     

基于Hough变换与最小二乘拟合的香烟小包拉线错牙检测方法

针对小盒香烟包装生产过程中经常出现外包装透明薄膜上没有拉线或拉线错牙等现象,提出一种基于图像处理技术的检测方法。首先,对烟包拉线图像进行边缘检测,获得拉线的轮廓;其次,利用Hough变换检测小盒拉线的上下边缘;当Hough变换无法完全找到拉线的上下边界时,采用最小二乘算法对左右两侧拉线弧进行圆拟合,最后根据计算出的拉线宽度或两圆心的垂直距离判断拉线是否错牙并计算出拉线错牙偏移量。实验结果表明所提方法能有效地检测出烟包拉线的错牙程度,且具有较好的鲁棒性和有效性。     

分类错误指导的分层B-CNN模型用于细粒度分类

目的 细粒度分类近年来受到了越来越多研究者的广泛关注,其难点是分类目标间的差异非常小。为此提出一种分类错误指导的分层双线性卷积神经网络模型。方法 该模型的核心思想是将双线性卷积神经网络算法（B-CNN）容易分错、混淆的类再分别进行重新训练和分类。首先,为得到易错类,提出分类错误指导的聚类算法。该算法基于受限拉普拉斯秩（CLR）聚类模型,其核心“关联矩阵”由“分类错误矩阵”构造。其次,以聚类结果为基础,构建了新的分层B-CNN模型。结果 用分类错误指导的分层B-CNN模型在CUB-200-2011、 FGVC-Aircraft-2013b和Stanford-cars 3个标准数据集上进行了实验,相比于单层的B-CNN模型,分类准确率分别由84.35%,83.56%,89.45%提高到了84.67%,84.11%,89.78%,验证了本文算法的有效性。结论 本文提出了用分类错误矩阵指导聚类从而进行重分类的方法,相对于基于特征相似度而构造的关联矩阵,分类错误矩阵直接针对分类问题,可以有效提高易混淆类的分类准确率。本文方法针对比较相近的目标,尤其是有非常相近的目标的情况,通过将容易分错、混淆的目标分组并进行再训练和重分类,使得分类效果更好,适用于细粒度分类问题。     

Event‐triggered distributed constrained consensus

We consider a distributed consensus problem for continuous‐time multi‐agent systems with set constraints on the final states. To save communication costs, an event‐triggered communication‐based protocol is proposed. By comparing its own instantaneous state with the one previously broadcasted to neighbours, each agent determines the next communication time. Based on this event‐triggered communication, each agent is not required to continuously monitor its neighbours' state and the communication only happens at discrete time instants. We show that, under some mild conditions, the constrained consensus of the multi‐agent system with the proposed protocol can be achieved with an exponential convergence rate. A lower bound of the transmission time intervals is provided that can be adjusted by choosing different values of parameters. Numerical examples illustrate the results. Copyright © 2016 John Wiley & Sons, Ltd.     

Reliability and time‐to‐failure bounds for discrete‐time constrained Markov jump linear systems

This paper presents a methodology to obtain a guaranteed‐reliability controller for constrained linear systems, which switch between different modes according to a Markov chain (Markov jump linear systems). Inside the classical maximal robust controllable set, there is 100% guarantee of never violating constraints at future time. However, outside such set, some sequences might make hitting constraints unavoidable for some disturbance realisations. A guaranteed‐reliability controller based on a greedy heuristic approach was proposed in an earlier work for disturbance‐free, robustly stabilisable Markov jump linear systems. Here, extensions are presented by, first, considering bounded disturbances and, second, presenting an iterative algorithm based on dynamic programming. In non‐stabilisable systems, reliability is zero; therefore, prior results cannot be applied; in this case, optimisation of a mean‐time‐to‐failure bound is proposed, via minor algorithm modifications. Optimality can be proved in the disturbance‐free, finitely generated case. Copyright © 2016 John Wiley & Sons, Ltd.     

基于B样条整体最小二乘的非线性多功能传感器信号重构方法

提出一种基于B样条整体最小二乘(Total least squares,TLS)的非线性多功能传感器信号重构新方法.该方法利用B样条基函数直接构建描述多功能传感器传递函数反函数的张量积B样条曲面;采用TLS求解超定方程组以获得稳定的控制系数估计.以二输入二输出多功能传感器模型为实验对象,在两种非线性情况下对多功能传感器的输入信号进行了重构,重构相对误差分别为0.162％和1.043％,并与常用重构方法进行了对比分析.理论和实验表明,B样条TLS重构方法对非线性多功能传感器传递函数的反函数具有良好的逼近性能,在信号重构中表现出较好的有效性.     

Sensorless supervision of linear dynamical systems: The Feed-Forward Command Governor approach

This paper proposes a novel class of Command Governor (CG) strategies for input and state-related constrained discrete-time LTI systems subject to bounded disturbances in the absence of explicit state or output measurements. While in traditional CG schemes the set-point manipulation is undertaken on the basis of either the actual measure of the state or its suitable estimation, it is shown here that the CG design problem can be solved, with limited performance degradation and with similar properties, also in the case that such an explicit measure is not available. This approach, which will be referred to as the Feed-Forward CG (FF-CG) approach, may be a convenient alternative CG solution in all situations whereby the cost of measuring the state may be a severe limitation, e.g. in distributed or decentralized applications. In order to evaluate the method proposed here, numerical simulations on a physical example have been undertaken and comparisons with the standard state-based CG solution reported.     

Complexity and design of QoS routing algorithms in wireless mesh networks

Quality of service (QoS) provisioning in wireless mesh networks (WMNs) is an open issue to support emerging multimedia services. In this paper, we study the problem of QoS provisioning in terms of end-to-end bandwidth allocation in WMNs. It is challenging due to interferences in the networks. We consider widely used interference models and show that except a few special cases, the problem of finding a feasible path is NP-complete under the models. We propose a k-shortest path based algorithmic framework to solve this problem. We also consider the problem of optimizing network performance by on-line dynamic routing, and adapt commonly used conventional QoS routing metrics to be used in WMNs. We find the optimal solutions for these problems through formulating them as optimization models. A model is developed to check the existence of a feasible path and another to find the optimal path for a demand; moreover, an on-line optimal QoS routing algorithm is developed. Comparing the algorithms implemented by the proposed framework with the optimization models shows that our solution can find existing feasible paths with high probability, efficiently optimizes path lengths, and has a comparable performance to the optimal QoS routing algorithm. Furthermore, our results show that contrary to wireline networks, minimizing resource consumption should be preferred over load distribution even in lightly loaded WMNs.     

A de-texturing and spatially constrained K-means approach for image segmentation

This paper presents a new and simple segmentation method based on the K-means clustering procedure and a two-step process. The first step relies on an original de-texturing procedure which aims at converting the input natural textured color image into a color image, without texture, that will be easier to segment. Once, this de-textured (color) image is estimated, a final segmentation is achieved by a spatially-constrained K-means segmentation. These spatial constraints help the iterative K-means labeling process to succeed in finding an accurate segmentation by taking into account the inherent spatial relationships and the presence of pre-estimated homogeneous textural regions in the input image. This procedure has been successfully applied on the Berkeley image database. The experiments reported in this paper demonstrate that the proposed method is efficient (in terms of visual evaluation and quantitative performance measures) and performs competitively compared to the best existing state-of-the-art segmentation methods recently proposed in the literature.