Michigan Particle Swarm Optimization for Prototype Reduction in Classification Problems

This paper presents a new approach to Particle Swarm Optimization, called Michigan Approach PSO (MPSO), and its application to continuous classification problems as a Nearest Prototype (NP) classifier. In Nearest Prototype classifiers, a collection of prototypes has to be found that accurately represents the input patterns. The classifier then assigns classes based on the nearest prototype in this collection. The MPSO algorithm is used to process training data to find those prototypes. In the MPSO algorithm each particle in a swarm represents a single prototype in the solution and it uses modified movement rules with particle competition and cooperation that ensure particle diversity. The proposed method is tested both with artificial problems and with real benchmark problems and compared with several algorithms of the same family. Results show that the particles are able to recognize clusters, find decision boundaries and reach stable situations that also retain adaptation potential. The MPSO algorithm is able to improve the accuracy of 1-NN classifiers, obtains results comparable to the best among other classifiers, and improves the accuracy reported in literature for one of the problems.

Hybrid crossover operators with multiple descendents for real‐coded genetic algorithms: Combining neighborhood‐based crossover operators

Most real‐coded genetic algorithm research has focused on developing effective crossover operators, and as a result, many different types of crossover operators have been proposed. Some forms of crossover operators are more suitable to tackle certain problems than others, even at the different stages of the genetic process in the same problem. For this reason, techniques that combine multiple crossovers, called hybrid crossover operators, have been suggested as alternative schemes to the common practice of applying only one crossover model to all the elements in the population. On the other hand, there are operators with multiple offsprings, more than two descendants from two parents, which present a better behavior than the operators with only two descendants, and achieve a good balance between exploration and exploitation. © 2009 Wiley Periodicals, Inc.     

Introducing requirements traceability support in model-driven development of web applications

In this work, we present an approach that introduces requirements traceability capabilities in the context of model-driven development of Web applications. This aspect allows us to define model-to-model transformations that not only provide a software artifact of lower abstraction (as model-to-model transformations usually do) but also to provide feedback about how they are applied. This feedback helps us to validate whether transformations are correctly applied. In particular, we present a model-to-model transformation that allows us to obtain navigational models of the Web engineering method OOWS from a requirements model. This transformation is defined as a set of mappings between these two models that have been implemented by means of graph transformations. The use of graph transformations allows us to develop a tool-supported strategy for applying mappings automatically. In addition, mechanisms for tracing requirements are also included in the definition of graph transformations. These mechanisms allow us to link each conceptual element to the requirements from which it is derived. In particular, we focus on tracing requirements throughout the navigational model, which describe the navigational structure of a Web application. To take advantage of these traceability mechanisms, we have developed a tool that obtains traceability reports after applying transformations. These reports help us to study aspects such as whether requirements are all supported, the impact of changing a requirement, or how requirements are modelled.     

Vision-based control for rigid body stabilization

This paper addresses the problem of stabilizing to a desired equilibrium point an eye-in-hand system, which consists of a single camera mounted on a rigid body free to move on . It is assumed that there is a collection of landmarks fixed in the environment and that the image coordinates of those landmarks are provided to the system by an on-board CCD camera. The proposed method addresses not only the problem of stabilization but also that of maintaining feature visibility along the system’s trajectory. The resulting solution consists of a feedback control law based on the current and desired image coordinates and reconstructed attitude and depth ratio information, which guarantees that (i) the desired equilibrium point is an almost global attractor; (ii) a set of necessary conditions for feature visibility holds throughout the system’s trajectories; and (iii) the image of a predefined feature point is kept inside the camera’s field of view.     

Robot arm reaching through neural inversions and reinforcement learning

We present a neural method that computes the inverse kinematics of any kind of robot manipulators, both redundant and non-redundant. Inverse kinematics solutions are obtained through the inversion of a neural network that has been previously trained to approximate the manipulator forward kinematics. The inversion provides difference vectors in the joint space from difference vectors in the workspace. Our differential inverse kinematics (DIV) approach can be viewed as a neural network implementation of the Jacobian transpose method for arm kinematic control that does not require previous knowledge of the arm forward kinematics. Redundancy can be exploited to obtain a special inverse kinematic solution that meets a particular constraint (e.g. joint limit avoidance) by inverting an additional neural network The usefulness of our DIV approach is further illustrated with sensor-based multilink manipulators that learn collision-free reaching motions in unknown environments. For this task, the neural controller has two modules: a reinforcement-based action generator (AG) and a DIV module that computes goal vectors in the joint space. The actions given by the AG are interpreted with regard to those goal vectors.     

3D posture estimation using geodesic distance maps

This paper presents a novel technique for three-dimensional (3D) human motion capture using a set of two non-calibrated cameras. The user’s five extremities (head, hands and feet) are extracted, labeled and tracked after silhouette segmentation. As they are the minimal number of points that can be used in order to enable whole body gestural interaction, we will henceforth refer to these features as crucial points. Features are subsequently labelled using 3D triangulation and inter-image tracking. The crucial point candidates are defined as the local maxima of the geodesic distance with respect to the center of gravity of the actor region that lie on the silhouette boundary. Due to its low computational complexity, the system can run at real-time paces on standard personal computers, with an average error rate range between 4% and 9% in realistic situations, depending on the context and segmentation quality.

Robot task plan representation by Petri nets: modelling, identification, analysis and execution

In this paper we introduce a framework to represent robot task plans based on Petri nets. Our approach enables modelling a robot task, analysing its qualitative and quantitative properties and using the Petri net representation for actual plan execution. The overall model is obtained from the composition of simple models, leading to a modular approach. Analysis is applied to a closed loop between the robot controller and the environment Petri net models. We focus here on the quantitative properties, captured by stochastic Petri net models. Furthermore, we introduce a method to identify the environment and action layer parameters of the stochastic Petri net models from real data, improving the significance of the model. The framework building blocks and a single-robot task model are detailed. Results of a case study with simulated soccer robots show the ability of the framework to provide a systematic modelling tool, and of determining, through well-known analysis methods for stochastic Petri nets, relevant properties of the task plan applied to a particular environment.