Nancy Arana-Daniel | Universidad de Guadalajara (original) (raw)

Papers by Nancy Arana-Daniel

Intelligent Data Analysis, Jul 13, 2016

Automatic cloud detection has played an important role in meteorological research. However, autom... more Automatic cloud detection has played an important role in meteorological research. However, automatically detecting clouds can also be useful in other fields such as aeronautics, especially for Unmanned Aerial Vehicles (UAVs), since going through dense clouds could destabilize the UAV. Also being aware of the clouds in the surroundings can decreasing the chances of a controlled flight into terrain scenario. This paper shows the design and development of several easy-to-implement superpixel segmentation descriptors with low computational cost, which are robust to incompleteness, geometric distortion, discrimination, and uniqueness. Four of the proposals are developed for cloud-sky classification, and a fifth proposal is made for ground-cloud-sky classification. Three of the approaches are generated from the extracted histograms of the superpixels obtained from the images. The fourth descriptor, used only for comparison, was obtained by applying SURF to superpixels.Our descriptors proposal is implemented for images obtained from video/photographic cameras mounted on a UAV. Due to its computational cost, it can be computed using low-performance computers.Experimental results showed that when using the proposed descriptors with a Support Vector Machine (SVM) classifier, the obtained recognition rates are improved in comparison with the state-of-the-art feature and texture descriptors used for cloud classification.

PeerJ Computer Science, 2021

This article presents an approach to solve the inverse kinematics of cooperative mobile manipulat... more This article presents an approach to solve the inverse kinematics of cooperative mobile manipulators for coordinate manipulation tasks. A self-adaptive differential evolution algorithm is used to solve the inverse kinematics as a global constrained optimization problem. A kinematics model of the cooperative mobile manipulators system is proposed, considering a system with two omnidirectional platform manipulators with n DOF. An objective function is formulated based on the forward kinematics equations. Consequently, the proposed approach does not suffer from singularities because it does not require the inversion of any Jacobian matrix. The design of the objective function also contains penalty functions to handle the joint limits constraints. Simulation experiments are performed to test the proposed approach for solving coordinate path tracking tasks. The solutions of the inverse kinematics show precise and accurate results. The experimental setup considers two mobile manipulators ...

Chapter 3 shows the design and implementation of a method used to reconstruct 3D surfaces from po... more Chapter 3 shows the design and implementation of a method used to reconstruct 3D surfaces from point-clouds, using Radial Basis Functions (RBFs) neural networks, which have been adjusted utilizing PSO algorithm. Meshing functions in order to interpolate point clouds to obtain compact surface representations is very important for Computer Aided Design (CAD), robot mapping, and object description and recognition, among other important applications. The results obtained using our algorithm show that, although the obtained surfaces are not continuous, they can be used as compact descriptors for pattern recognition process, and environmental mapping. This is because our proposal is fast enough to be implemented in real time, and the reduction of the number of parameters used to describe a shape with 3D point clouds is significant.

2018 IEEE Latin American Conference on Computational Intelligence (LA-CCI), 2018

The solution of the inverse kinematics of robot manipulators may have multiple joints configurati... more The solution of the inverse kinematics of robot manipulators may have multiple joints configurations that reach the same end effector pose. These redundant solutions may provide a collision free joint configuration for a safe inverse kinematics task. In this work, we propose to solve the inverse kinematics problem for robot manipulators using the Firefly Algorithm. The proposed method is able to provide one or more solutions for the given inverse kinematics task. Since the method does not required the inversion of any Jacobian matrix, it avoids singularities configurations. Applicability of the proposed approach is illustrated in simulations using a five degree of freedom arm manipulator.

This work presents a scheme based on a discrete recurrent high order neural network identifier an... more This work presents a scheme based on a discrete recurrent high order neural network identifier and a block control based on sliding modes for nonlinear discrete-time systems with input delays in real-time. The identifier is trained with an extended Kalman Filter based algorithm and the block control is used for trajectory tracking. Experimental results are included using a linear induction motor prototype with added delays to its input signals.

Chapter 6 presents the application of bio-inspired algorithms to improve neural identifiers for d... more Chapter 6 presents the application of bio-inspired algorithms to improve neural identifiers for discrete-time unknown nonlinear systems. PSO is particularly used to improve two kinds of neural identifiers: first, PSO is used to find initial conditions of an EKF learning algorithm (enhanced PSO-EKF) to train a RHONN in order to identify a dynamic mathematical model of a linear induction motor benchmark; second, the enhanced PSO-EKF is used to train a recurrent multilayer perceptron in order to obtain an accurate neural model for forecasting in smart grids. Importance of these applications is attributable to the need of accurate dynamic models for modern purposes, such as control, forecasting, simulation, and emulation. In addition to the foregoing applications the PSO-EKF combination has shown its applicability for the proposed schemes for different kinds of unknown nonlinear-systems with noises, uncertainties, delays, saturations, etc.

2017 International Joint Conference on Neural Networks (IJCNN), 2017

In recent years, the research on neural networks has been guided by the search of new mathematica... more In recent years, the research on neural networks has been guided by the search of new mathematical frameworks, with the hope of finding new features, as geometric interpretation, for facing today problems or reducing the computational cost. In this paper we introduce a new Clifford Neuron [1], extending the conformai neuron, presented in [2] through the generalization of the geometric algebra of quadratic surfaces (G6,3), presented in [3]. In this new neuron, we can obtain decision surfaces with different geometric shapes, depending on the input data: spherical decision surface, ellipsoidal, cylindrical or even decision surface as a pair of planes (all of them can be derived as special case of an ellipse). The above without the need of using a kernel technique, just using a linear activation function over the hiperconformal space.

This paper presents an approach to perform local navigation in outdoor environments using a bio-i... more This paper presents an approach to perform local navigation in outdoor environments using a bio-inspired algorithm. The proposed approach uses the Particle Swarm Optimization (PSO) to perform the robot navigation. The PSO particles represent a possible new position in the navigation task. The best PSO particle is chosen and is transformed into latitude and longitude values. Finally, given the desired latitude and longitude values a controller is used to move the robot from its current position and orientation to the valid and best PSO particle in each iteration until reaching the goal given in latitude and longitude.

Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 2019

Nowadays, the Adam algorithm has become one of the most popular optimizers to train feed-forward ... more Nowadays, the Adam algorithm has become one of the most popular optimizers to train feed-forward neural networks because it takes the best features of other gradient-based optimizers, such as working well with sparse gradients, in online and non-stationary settings, and also it is very robust to the rescaling of the gradient. The above makes Adam the best choice to solve problems with non-stationary objectives, very noise gradients, and with large data inputs. In this work, we enhanced the Adam algorithm by using the Kalman filter, and the novel proposal is called KAdam. Instead of using the computed gradients directly from the cost function, we first apply the Kalman filter on them. As a result, the filtered gradients allow the algorithm to explore new (and potentially better) solutions on the cost function. The results obtained when applying our proposal and other state-of-the-art optimizers to solve classification problems show that KAdam is able to obtain better accuracies than its competitors in the same execution time.

Integrated Computer-Aided Engineering, 2021

The inverse kinematics of robotic manipulators consists of finding a joint configuration to reach... more The inverse kinematics of robotic manipulators consists of finding a joint configuration to reach a desired end-effector pose. Since inverse kinematics is a complex non-linear problem with redundant solutions, sophisticated optimization techniques are often required to solve this problem; a possible solution can be found in metaheuristic algorithms. In this work, a modified version of the firefly algorithm for multimodal optimization is proposed to solve the inverse kinematics. This modified version can provide multiple joint configurations leading to the same end-effector pose, improving the classic firefly algorithm performance. Moreover, the proposed approach avoids singularities because it does not require any Jacobian matrix inversion, which is the main problem of conventional approaches. The proposed approach can be implemented in robotic manipulators composed of revolute or prismatic joints of n degrees of freedom considering joint limits constrains. Simulations with differen...

Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 2018

2018 International Joint Conference on Neural Networks (IJCNN), 2018

Infectious diseases are causes of morbidity and mortality worldwide. Mathematical models can serv... more Infectious diseases are causes of morbidity and mortality worldwide. Mathematical models can serve as a central tool to predict the kinetic of different infections. However, the development of mechanistic models and their parameter estimation are difficult tasks. Using Recurrent High Order Neural Networks (RHONNs) trained with an algorithm based on the extended Kalman filter (EKF), we separately identified influenza A virus (IAV) and HIV dynamics. To this end, we considered within-host mathematical models of IAV and HIV as unknown signals to the RHONNs. Simulations results reported that for both infections, RHONNs are able to identify the within-host model dynamics. Results provide promising guidelines to tackle the problem of model identification of infectious diseases, serving for future model based control strategies of viral infections.

2017 IEEE Latin American Conference on Computational Intelligence (LA-CCI), 2017

This paper deals with parametric identification for discrete-time α-ß model for three phase linea... more This paper deals with parametric identification for discrete-time α-ß model for three phase linear induction motors (LIM). This parametric identification is performed using the well-known PSO algorithm, using experimental data obtained from a real-time implementation on a LIM benchmark. Obtained parameters are validated using signal fitting for state variables, under presence of unknown disturbances and time-delays.

Intelligent Data Analysis, 2020

This paper presents an improved extension of the previous algorithm of the authors called KAdam t... more This paper presents an improved extension of the previous algorithm of the authors called KAdam that was proposed as a combination of a first-order gradient-based optimizer of stochastic functions, known as the Adam algorithm and the Kalman filter. In the extension presented here, it is proposed to filter each parameter of the objective function using a 1-D Kalman filter; this allows us to switch from matrix and vector calculations to scalar operations. Moreover, it is reduced the impact of the measurement noise factor from the Kalman filter by using an exponential decay in function of the number of epochs for the training. Therefore in this paper, is introduced our proposed method sKAdam, a straightforward improvement over the original algorithm. This extension of KAdam presents a reduced execution time, a reduced computational complexity, and better accuracy as well as keep the properties from Adam of being well suited for problems with large datasets and/or parameters, non-statio...

2019 IEEE International Autumn Meeting on Power, Electronics and Computing (ROPEC), 2019

Long Short-Term Memory (LSTM) is a type of recurrent neural network that has become important in ... more Long Short-Term Memory (LSTM) is a type of recurrent neural network that has become important in machine learning research thanks to its high precision to solve problems such as speech recognition, handwriting recognition, natural text compression, sequential data processing among others. Although classic LSTM are powerful tools to solve such problems, their adaptation is far from showing a smooth behavior which represents a drawback to LSTM be used in applications such as real-time control of physical systems in which to fulfill restrictions of ranges of values of the control variables is important in order to preserve the physical integrity of the systems. In this paper we present a design of architecture of LSTM that overcomes the non-smooth adaptation problem by using a single forget gate for all the LSTM units and furthermore improves the accuracy of classic LSTMs in problems such as rebber grammar learning, time series forecasting and control of physical systems as it is shown...

Chapter 7 presents the use of bio-inspired algorithms to improve neural controllers for discrete-... more Chapter 7 presents the use of bio-inspired algorithms to improve neural controllers for discrete-time unknown nonlinear systems utilizing two approaches. First, a Neural-PSO Second-Order Sliding Mode Controller approach is applied to control a class of unknown nonlinear systems; second, a Neural-BFO Second-Order Sliding Mode Controller is incorporated into the same class of unknown nonlinear systems. In order to show applicability of these controllers, they are applied to a Van der Pol Oscillator and a comparative analysis is undertaken to establish conclusions about the development of both controllers with respect to a traditional one. The Neural-BFO Second-Order Sliding Mode Controller proved to perform better in the trajectory tracking of a class of unknown discrete-time nonlinear systems with disturbances (external an internal).

Chapter 5 deals with a soft computing approach that is able to avoid obstacles while moving a rob... more Chapter 5 deals with a soft computing approach that is able to avoid obstacles while moving a robot to reach a goal. The approach is based on the PSO algorithm, where each particle represents a potential solution of a new position for the robot. Once the best particle of the actual iteration is selected, the robot is moved to the position that the best particle represents. This algorithm is tested with nonholonomic and holonomic robots and proves that bio-inspired algorithms are able to solve local navigation problems in real-time.

Electronics, 2020

In this paper, an adaptive single neuron Proportional–Integral–Derivative (PID) controller based ... more In this paper, an adaptive single neuron Proportional–Integral–Derivative (PID) controller based on the extended Kalman filter (EKF) training algorithm is proposed. The use of EKF training allows online training with faster learning and convergence speeds than backpropagation training method. Moreover, the propose adaptive PID approach includes a back-calculation anti-windup scheme to deal with windup effects, which is a common problem in PID controllers. The performance of the proposed approach is shown by presenting both simulation and experimental tests, giving results that are comparable to similar and more complex implementations. Tests are performed for a four wheeled omnidirectional mobile robot. Tests show the superiority of the proposed adaptive PID controller over the conventional PID and other adaptive neural PID approaches. Experimental tests are performed on a KUKA® Youbot® omnidirectional platform.