Fernando Martin

Medical Data Analysis

In its first part, this contribution reviews shortly the application of neural network methods to medical problems and characterizes its advantages and problems in the context of the medical background. Successful application examples show that human diagnostic capabilities are significantly worse than the neural diagnostic systems. Then, paradigm of neural networks is shortly introduced and the main problems of medical data base and the basic approaches for training and testing a network by medical data are described. Additionally, the problem of interfacing the network and its result is given and the neurofuzzy approach is presented. Finally, as case study of neural rule based diagnosis septic shock diagnosis is described, on one hand by a growing neural network and on the other hand by a rule based system.

Path Planning for Mobile Robot Navigation using Voronoi Diagram and Fast Marching

This paper presents a new sensor based global path planner which operates in two steps. In the first step the safest areas in the environment are extracted by means of a Voronoi diagram. In the second step fast marching method is applied to the Voronoi extracted areas in order to obtain the shortest path. In this way the trajectory obtained is the shortest between the safe possible ones. This two step method combines an extremely fast global planner operating on a simple sensor based environment modeling, while it operates at the sensor frequency. The main characteristics are speed and reliability, because the map dimensions are reduced to a unidimensional map and this map represents the safest areas in the environment for moving the robot

DE-based tuning of PI(λ)D(μ) controllers.

A new method that relies on evolutionary computation concepts is proposed in this paper to tune the parameters of fractional order PI(λ)D(μ) controllers, in which the orders of the integral and derivative parts, λ and μ, respectively, are fractional. The main advantage of the fractional order controllers is that the increase in the number of parameters in the controller allows an increase in the number of control specifications that can be met. A Differential Evolution (DE) algorithm is proposed to make the controlled system fulfill different design specifications in time and frequency domains. This method is based on the minimization of a fitness function. Experiments have been carried out in simulation and in a real DC motor platform. The results illustrate the effectiveness of this method.

Kullback-Leibler divergence-based global localization for mobile robots

The global localization problem for mobile robots is addressed in this paper. In this field, the most common approaches solve this problem based on the minimization of a quadratic loss function or the maximization of a probability distribution. The distances obtained from the perceptive sensors are used together with the predicted ones (from the estimates in the known map) to define a cost function or a probability to optimize. In our previous work, we developed an optimization-based global localization module that used evolutionary computation concepts. In particular, the algorithm engine was the Differential Evolution method. In this work, this algorithm has been modified including the minimization of the Kullback-Leibler divergence between true observations and estimates. This divergence is used to calculate the cost function of the localization module. The algorithm has been tested in different situations and the most important improvement is the ability to cope with different types of occlusions.

High-accuracy global localization filter for three-dimensional environments

The localization problem in mobile robotics can be defined as the search of the robots coordinates in a known environment. If there is no information about the initial location, we are talking about global localization. In this work, we have developed an algorithm that solves this problem in a three-dimensional (3D) environment using evolutionary computation concepts. The method has been called RELF-3D and has many features that make it very robust and reliable: thresholding and discarding mechanisms, different cost functions, effective convergence criteria, and so on. The resulting global localization module has been tested in numerous experiments and the most important improvement obtained is the accuracy of the method, allowing its application in manipulation tasks.

Smooth path planning for non-holonomic robots using fast marching

This paper presents the application to nonholonomic mobile robot path planning of our Voronoi Fast Marching (VFM) and FM2 methods, which represents our current progress on the design and analysis of these algorithms. The VFM and FM2 methods use the propagation of a wave (Fast Marching) operating on the world model, to determine a motion plan over a slowness map (similar to the refraction index in Optics) extracted from the updated map model. The computational efficiency of the method let the planner operate at high rate sensor frequencies. This method allow us to simplify the mobile robot or mobile manipulator architecture, while maintaining good response time and smooth and safe planned trajectories. This method can be classified inside the navigation functions (a type of potential fields) and it is complete (it finds the solution path if it exists) and of order n complexity (O(n)). The results presented in the paper show how the proposed method is faster than other existing path planning methods for non-holonomic (car like for example) mobile robots and generates trajectories of better quality .

A New Method for Unifying Heterogeneous Databases

In this article we present a new method for unifying heterogeneous databases. We propose the unification at the level of conceptual schemas in a two-step process. First, the internal database schemas are mapped to conceptual schemas in a user interactive process. These schemas are then automatically integrated, creating a virtual repository. Virtual repositories give users the feeling of working with single, local databases and can be used for epidemiological research (e.g. data analysis) or clinical practice. We implemented the proposed method in a tool for accessing medical information from remote databases located at different machines with different technological platforms in the Internet.

Differential Evolution Markov Chain Filter for Global Localization

A key challenge for an autonomous mobile robot is to estimate its location according to the available information. A particular aspect of this task is the global localization problem. In our previous work, we developed an algorithm based on the Differential Evolution method that solves this problem in 2D and 3D environments. The robot’s pose is represented by a set of possible location estimates weighted by a fitness function. The Markov Chain Monte Carlo algorithms have been successfully applied to multiple fields such as econometrics or computing science. It has been demonstrated that they can be combined with the Differential Evolution method to solve efficiently many optimization problems. In this work, we have combined both approaches to develop a global localization filter. The algorithm performance has been tested in simulated and real maps. The population requirements have been reduced when compared to the previous version.

Log of the inverse of the Distance Transform and Fast Marching applied to Path Planning

This paper presents a new path planning method based in the inverse of the logarithm of the distance transform and in the fast marching method. The distance transform of an image gives a grey scale that is darker near the obstacles and walls and more clear far from them and it is calculated via Voronoi diagram. The logarithm of the inverse of the distance transform imitates the repulsive electric potential from walls and obstacles. This method is very fast and reliable and the trajectories are similar to the human trajectories: smooth and not very close to obstacles and walls

Automatic detection and verification of pipeline construction features with multi-modal data

Assessment of the condition of underground pipelines is crucial to avoid breakages. Autonomous in-line inspection tools provided with Non-destructive Technology (NDT) sensors to assess large sections of the pipeline are commonly used for these purposes. An example of such sensors based on Eddy currents is the Remote Field Technology (RFT). A crucial step during in-line inspections is the detection of construction features, such as joints and elbows, to accurately locate and size specific defects within pipe sections. This step is often performed manually with the aid of visual data, which results in slow data processing. In this paper, we propose a generic framework to automate the detection and verification of these construction features using both NDT sensor data and visual images. Firstly, supervised learning is used to identify the construction features in the NDT sensor signals. Then, image processing is employed to verify the selection. Results are presented with data from a RFT tool, for which a specialised descriptor has been designed to characterise and classify its signal features. Furthermore, the construction feature is displayed in the image, once it is identified in the RFT data and detected in the visual data. A visual odometry algorithm has been implemented to locate the visual data with respect to the RFT data. About 800 meters of these multi-modal data are evaluated to test the validity of the proposed approach.