Showing papers by "Xiaoou Li published in 2012"

Anti-Swing Control For An Overhead Crane With Fuzzy Compensation

Abstract: This paper proposes a novel anti-swing control strategy for an overhead crane. The controller includes both position regulation and anti-swing control. Since the crane model is not exactly known, fuzzy rules are used to compensate friction, gravity as well as the coupling between position and anti-swing control. Ahigh-gain observer is introduced to estimate the joint velocities to realize PD control. Real-time experiments are presented comparing this new stable anti-swing PID control strategy with regular crane controllers.

Fast classification for large data sets via random selection clustering and Support

Abstract: Support Vector Machines (SVMs) are high-accuracy classifiers. However, normal SVM algorithms are unsuitable for classification of large data sets because of their training complexity. In this paper, we propose a novel SVM classification approach for large data sets. We first use the random selection to select a small group of training data for the first-stage SVM. Then a de-clustering technique is proposed to recover the training data for the second-stage SVM. This two-stage SVM classifier has distinctive advantages on dealing with huge data sets such as those in bioinformatics. The performance analysis is also given in this paper. Finally, we apply the proposed method on several benchmark problems. Experimental results demonstrate that this approach has good classification accuracy while the training is significantly faster than other SVM classifiers.

Haar wavelet neural networks for nonlinear system identification

Abstract: Since wavelet transform uses the multi-scale (or multi-resolution) techniques for time series, wavelet transform is suitable for modeling complex signals. Haar wavelet transform is the most commonly used and the simplest one. The Haar wavelet neural network (HWNN) applies the Harr wavelet transform as active functions. It is easy for HWNN to model a nonlinear system at multiple time scales and sudden transitions. In this paper, two types of HWNN, feedforward and recurrent wavelet neural networks, are used to model discrete-time nonlinear systems, which are in the forms of the NARMAX model and state-space model. We first propose an optimal method to determine the structure of HWNN. Then two stable learning algorithms are given for the shifting and broadening coefficients of the wavelet functions. The stability of the identification procedures is proven.

Fast classification for large data sets via random selection clustering and Support Vector Machines

Abstract: Support Vector Machines SVMs are high-accuracy classifiers. However, normal SVM algorithms are unsuitable for classification of large data sets because of their training complexity. In this paper, we propose a novel SVM classification approach for large data sets. We first use the random selection to select a small group of training data for the first-stage SVM. Then a de-clustering technique is proposed to recover the training data for the second-stage SVM. This two-stage SVM classifier has distinctive advantages on dealing with huge data sets such as those in bioinformatics. The performance analysis is also given in this paper. Finally, we apply the proposed method on several benchmark problems. Experimental results demonstrate that this approach has good classification accuracy while the training is significantly faster than other SVM classifiers.

Convex-Concave Hull for Classification with Support Vector Machine

Abstract: Support vector machine (SVM) is not suitable for classification on large data sets due to its training complexity. Convex hull can simplify SVM training, however the classification accuracy becomes lower when there are inseparable points. This paper introduces a novel method for SVM classification, called convex-concave hull. After a grid processing, the convex hull is used to find extreme points. Then we detect a concave (non-convex) hull, the vertices of it are used to train SVM. We applied the proposed method on several problems. Experimental results demonstrate that our approach has good classification accuracy while the training is significantly faster than other state of the art methods.

Data Selection Using Decision Tree for SVM Classification

Abstract: Support Vector Machine (SVM) is an important classification method used in a many areas. The training of SVM is almost O(n^{2}) in time and space. Some methods to reduce the training complexity have been proposed in last years. Data selection methods for SVM select most important examples from training data sets to improve its training time. This paper introduces a novel data reduction method that works detecting clusters and then selects some examples from them. Different from other state of the art algorithms, the novel method uses a decision tree to form partitions that are treated as clusters, and then executes a guided random selection of examples. The clusters discovered by a decision tree can be linearly separable, taking advantage of the Eidelheit separation theorem, it is possible to reduce the size of training sets by carefully selecting examples from training sets. The novel method was compared with LibSVM using public available data sets, experiments demonstrate an important reduction of the size of training sets whereas showing only a slight decreasing in the accuracy of classifier.

Structural Health Monitoring of Tall Buildings with Numerical Integrator and Convex-Concave Hull Classification

Abstract: An important objective of health monitoring systems for tall buildings is to diagnose the state of the building and to evaluate its possible damage. In this paper, we use our prototype to evaluate our data-mining approach for the fault monitoring. The offset cancellation and high-pass filtering techniques are combined effectively to solve common problems in numerical integration of acceleration signals in real-time applications. The integration accuracy is improved compared with other numerical integrators. Then we introduce a novel method for support vector machine (SVM) classification, called convex-concave hull. We use the Jarvis march method to decide the concave (nonconvex) hull for the inseparable points. Finally the vertices of the convex-concave hull are applied for SVM training.

Stable PID control for robot manipulators with neural compensation

Abstract: In order to minimize steady-state error with respect to uncertainties in robot control, the integral gain of PID control should be increased. Another method is to add a compensator to PD control, such as neural compensator, but the derivative gain of this PD control should be large enough. These two approaches deteriorate transient performances. In this paper, the popular neural PD is extended to neural PID control. The semiglobal asymptotic stability of the neural PID control is proven. The conditions give explicit selection methods for the gains of the linear PID control. A experimental study on an upper limb exoskeleton with this neural PID control is addressed.

DYMOND: an active system for dynamic vertical partitioning of multimedia databases

Abstract: In recent years, vertical partitioning techniques have been employed in multimedia databases to achieve efficient retrieval of multimedia objects. These techniques are static because the input to the partitioning process, which includes queries accessing database and their frequency as well as the database schema, is obtained from an earlier analysis stage. This implies that when the system undergoes sufficient changes, a new analysis stage is carried out to re-run the partitioning process. Multimedia databases are accessed by many users simultaneously, therefore queries and their frequency tend to quickly change over time. In this context, dynamic vertical partitioning can significantly improve performance. In this paper we present an active system called DYMOND (DYnamic Multimedia ON line Distribution), which performs a dynamic vertical partitioning in multimedia databases to improve query performance. Experimental results on benchmark multimedia databases clarify the validness of our system.

Robot PD control with parallel/serial neural network and sliding mode compensations

Abstract: Both neural network and sliding mode can compensate the steady-state error of proportional-derivative (PD) control. PD control with neural compensation is smooth, but it is not asymptotically stable. PD control with sliding mode is asymptotically stable, but the chattering is big. This paper first analyzes the asymptotic stability of PD control with parallel neural networks and the first-order sliding mode compensation. Then a serial compensation structure is proposed. In the serial compensation, a dead-zone neural PD control assures that the regulation error is bounded. And a super-twisting second-order sliding-mode is used to guarantee finite time convergence of the sliding mode PD control.

Dynamic Vertical Partitioning of Multimedia Databases Using Active Rules

Abstract: Vertical partitioning is a design technique widely employed in relational databases to reduce the number of irrelevant attributes accessed by the queries. Currently, due to the popularity of multimedia applications on the Internet, the need of using partitioning techniques in multimedia databases has arisen in order to use their potential advantages with regard to query optimization. In multimedia databases, the attributes tend to be of very large multimedia objects. Therefore, the reduction in the number of accesses to irrelevant objects would imply a considerable cost saving in the query execution. Nevertheless, the use of vertical partitioning techniques in multimedia databases implies two problems: 1) most vertical partitioning algorithms only take into account alphanumeric data, and 2) the partitioning process is carried out in a static way. In order to address these problems, we propose an active system called DYMOND, which performs a dynamic vertical partitioning in multimedia databases to improve query performance. Experimental results on benchmark multimedia databases clarify the validness of our system.

Maximally permissive control of flexible manufacturing systems with fewest disjunctive constraints

Abstract: For Petri net models whose legal reachability spaces are non-convex, one cannot optimally control them by the conjunctions of linear constraints. This work proposes a method to find a set of linear constraints such that each constraint can forbid all first-met bad markings and every legal marking satisfies at least one constraint. Then, the disjunctions of the obtained constraints can make all legal markings reachable and forbid all first-met bad markings, i.e., the controlled net is live and maximally permissive. The work also develops an integer linear programming model to minimize the number of the constraints. Finally, an example is provided to illustrate the proposed method.