R. J. Duro

Computational Intelligence for Remote Sensing

This book is a composition of different points of view regarding the application of Computational Intelligence techniques and methods to Remote Sensing data and applications. It is the general consensus that classification, its related data processing, and global optimization methods are core topics of Computational Intelligence. Much of the content of the book is devoted to image segmentation and recognition, using diverse tools from different areas of the Computational Intelligence field, ranging from Artificial Neural Networks to Markov Random Field modeling. The book covers a broad range of topics, starting from the hardware design of hyperspectral sensors, and data handling problems, namely data compression and watermarking issues, as well as autonomous web services. The main contents of the book are devoted to image analysis and efficient (parallel) implementations of these analysis techniques. The classes of images dealt with throughout the book are mostly multispectral-hyperspectral images, though there are some instances of processing Synthetic Aperture Radar images.

Biologically inspired robot behavior engineering

1. Evolutionary approaches to neural control of rolling, walking, swimming and flying animats or robots.- 2. Behavior coordination and its modification on monkey-type mobile robot.- 3. Visuomotor control in flies and behavior-based agents.- 4. Using evolutionary methods to parameterize neural models: a study of the lamprey central pattern generator.- 5. Biologically inspired neural network approaches to real-time collision-free robot motion planning.- 6. Self-adapting neural networks for mobile robots.- 7. Evolving robots able to integrate sensory-motor information over time.- 8. A non-computationally-intensive neurocontroller for autonomous mobile robot navigation.- 9. Some approaches for reusing behaviour based robot cognitive architectures obtained through evolution.- 10. Modular neural architectures for robotics.- 11. Designing neural control architectures for an autonomous robot using vision to solve complex learning tasks.- 12. Robust estimation of the optical flow based on VQ-BF.- 13. Steps towards one-shot vision-based self-localization.- 14. Computing the optimal trajectory of arm movement: the TOPS (Task Optimization in the Presence of Signal-dependent noise) model.- 15. A general learning approach to visually guided 3D-positioning and pose control of robot arms.

Considerations in the application of evolution to the generation of robot controllers

Abstract This paper is concerned with different aspects of the use of evolution for the successful generation of real robot Artificial Neural Network (ANN) controllers. Several parameters of an evolutionary/genetic algorithm (GA) and the way they influence the evolution of ANN behavioral controllers for real robots have been contemplated. These parameters include the way the initial populations are distributed, how the individuals are evaluated, the implementation of race schemes, etc. A batch of experiments on the evolution of three types of behaviors with different population sizes have been carried out in order to ascertain their effect on the evolution of the controllers and their validity in real implementations. The results provide a guide to the design of evolutionary algorithms for generating ANN based robot controllers, especially when, due to computational constraints, the populations to be used are small with respect to the complexity of the problem to be solved. The problem of transferring the controllers evolved in simulated environments to the real systems operating in real environments are also considered and we present results of this transference to reality with a robot which has few and extremely noisy sensors.

Evolutionary generation and training of recurrent artificial neural networks

An evolutionary artificial neural network training and design methodology is presented, aimed at obtaining optimum or quasi-optimum synchronous recurrent neural networks capable of processing sequential inputs. We show that, through the use of this method and working with floating point and integer valued chromosomes, it is possible to achieve optimum results, considering very small populations and few generations. In order to implement this methodology, we have developed GENIAL, a genetic algorithm development environment which is specifically designed for solving this type of problem. It offers ways of testing adequate fitness functions and many tools for improving results. Finally, we comment on the sequential introduction of different constraints in genetic algorithms, presenting a classical example where several design requirements are met simultaneously and which demonstrates the power of this method.<<ETX>>

Applying synaptic delays for virtual sensing and actuation in mobile robots

In this article we describe the use of Artificial Neural Networks (ANN) with synaptic time delays between the nodes as a means to increase the capabilities of the usual control modules used in behavior based robotics. This inclusion allows the controllers to manage explicit temporal information in different levels. In the sensing level it permits the use of virtual sensors that improve the precision of the information provided by sensors through a temporal correlation of their values. In the actuation level we use the network with an infrasensorized robot in a problem that requires active sensing, where the control and actuation mechanisms are coordinated in order to obtain a better sensorial image of the environment by means of a spatio-temporal representation of a perception sequence. The decision of the appropriate delays is automated through learning and evolution.

Synaptic delay based artificial neural networks and discrete time backpropagation applied to QRS complex detection

In this paper we make use of an extension of the backpropagation algorithm to discrete time feedforward networks that include internal time delays in the synapses. The structure of the network is similar to the one presented by Day-Davenport (1993), that is, in addition to the weights of the synaptic connections, we model their length through a parameter that indicates the delay a discrete event suffers when going from the origin neuron to the target neuron through a synaptic connection. Like the weights, these delays are also trainable, and a training algorithm can be obtained that is almost as simple as the backpropagation algorithm, and which is really an extension of it. We present an application of these networks to the task of identifying normal QRS and ventricular QRS complexes in an ECG signal with the network receiving the signal sequentially, that is, no windowing or segmentation is applied.

Robust visual recognition with high-order Gaussian synapses networks

In the context of visual systems for robots, we have made use of a high order gaussian synapses network and the Gaussian Synapses Backpropagation Algorithm (GSBP) for the implementation of the detectors that constitute one part of the whole visual architecture. These detectors are trained to be sensitive to spatial patterns that are relevant for the decisions the robot must perform during its operation in an environment. The inclusion of gaussian functions in the synapses of the network allows the network to select the appropriate spatial information and filter out all that is irrelevant according to the training it has received. In this paper we will show how these networks are easily trained to ignore backgrounds. In addition, with a very simple training set and an appropriate input selection strategy, the networks detect objects independently of size and position. These systems, coupled with an attention mechanism result in a very efficient visual information processor.

Discrete Time Backpropagation and Synaptic Delay Based Artificial Neural Networks in Chaotic Time Series Prediction

This paper is concerned with the application of a new training algorithm for delay based neural networks to the prediction of future values in chaotic time series. In the networks we employ, the transmission of information through synapses is delayed by a trainable amount. The main application of these structures is in training to perform operations that require reasoning with events occurring in different instants of time without any time windowing process. We test the validity of the approach to the prediction of future values in chaotic time series using iterative multistep prediction.

Design of ANN architectures for handling the temporal dimension in signal processing

We consider in this work two Artificial Neural Network architectures for the processing of time varying signals without resorting to a windowing process. They correspond to the two types of approaches one may take for this task. These architectures must have the capacity of handling the temporal dimension in their structure, avoiding the problems of the use of static architectures. The first topology is based on an extension of the concept of Finite State Automaton to cover a continuum of states and possible outputs. The second one is an ANN design that includes delays in the synaptic connections between neurons, thus further modeling the behavior of real neurons. We carry out a study of the characteristics and design needs of both approaches, and we apply them to QRS detection and identification in Electrocardiographic signals as an example of a real practical application.