Marconi Kolm Madrid

A simulator using classifier systems with neural networks for autonomous robot navigation

This paper presents a simulator that was developed to assist in the process of implementing high-level autonomous robot navigation algorithms and in the related experimentations. The classifier systems are designed using neural networks as classifiers to perform autonomous navigation. We propose a powerful simulator using classes and objects to be easily updated and extended. The simulator carries a class composed of methods for differential wheels steering, collision detection and sensor readings. Another class allows the specification of geometric shaped objects, which can also be detected as obstacles in the environment. In addition, operators are available to deal with credit assignment, genetic algorithms, and inference of the classifiers. By designing and constructing the simulator, we create conditions to explore the potentialities of neural networks as classifiers.

Characterization of multiscroll attractors using Lyapunov exponents and Lagrangian coherent structures.

The present work aims to apply a recently proposed method for estimating Lyapunov exponents to characterize-with the aid of the metric entropy and the fractal dimension-the degree of information and the topological structure associated with multiscroll attractors. In particular, the employed methodology offers the possibility of obtaining the whole Lyapunov spectrum directly from the state equations without employing any linearization procedure or time series-based analysis. As a main result, the predictability and the complexity associated with the phase trajectory were quantified as the number of scrolls are progressively increased for a particular piecewise linear model. In general, it is shown here that the trajectory tends to increase its complexity and unpredictability following an exponential behaviour with the addition of scrolls towards to an upper bound limit, except for some degenerated situations where a non-uniform grid of scrolls is attained. Moreover, the approach employed here also provides an easy way for estimating the finite time Lyapunov exponents of the dynamics and, consequently, the Lagrangian coherent structures for the vector field. These structures are particularly important to understand the stretching/folding behaviour underlying the chaotic multiscroll structure and can provide a better insight of phase space partition and exploration as new scrolls are progressively added to the attractor.

A complement to the back-propagation algorithm: an upper bound for the learning rate

This paper presents as complement for the back-propagation algorithm, a local convergence analysis derived from a vectorial approach. The analysis is made through application of Lyapunovs second method, and it supplies an upper bound for the learning rate.

Information Generation and Lagrangian Coherent Structures in Multiscroll Attractors

Abstract The present work aims to apply a recently alternative approach to estimate the Lyapunov exponents that is suitable for such piecewise linear model in order of characterizing - with the aid of the metric entropy - the degree of information generated by several kinds of multiscroll attractors. Another analytical path is followed with the extraction of the corresponding Lagrangian coherent structures, which allows the determination of the dynamical system separatrices and also of the folding/stretching behavior that results in chaos. Overall, the reported study can be understood as an investigation, from distinct theoretical standpoints, of key aspects regarding the dynamic behavior of systems capable of engendering multiscroll structures.

Recursive algorithm for the inverse kinematics of redundant robotic manipulators

Abstract This paper presents an alternative method for the inverse kinematics problem in serial-chain redundant robots. Such method is based on a recursive algorithm that solves inverse kinematics by allowing only one joint to move at a time, in a simulated stage. This transforms the n -dimensional problem in simpler unidimensional ones, whose analytical solution for each joint is presented using the Denavit-Hartenberg representation. Matlab simulations are performed in order to show the methods efficiency. The proposed method easily handles limitations of joint positions and velocities and can efficiently be applied in real time.

Identification of the Dynamical Model in Robotic Systems Using Only Information About the Position

Abstract This paper considers the identification of the dynamical model for robotic systems in which only information about the position of joints is available. In this condition, several robotic systems become unobservable from the linear viewpoint. A multivariable nonlinear identifier model based on the generic observability property of nonlinear systems is used to identify the dynamical model, and the multi-layer feedforward perceptron (MLP) class of artificial neural networks (ANN’s) is used as a tool for the nonlinear mathematic modeling.