Pablo Soto-Quiros
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Pablo Soto-Quiros has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Anatoli Torokhti
University of South Australia, Australia
9 shared publications - Luis Alberto Chavarría-Zamora
Instituto Tecnológico de Costa Rica
3 shared publications - Phil Howlett
University of South Australia
3 shared publications - Jeffry Chavarría-Molina
Escuela de Matemática, Instituto Tecnológico de Costa Rica, Cartago, Costa Rica
2 shared publications - Juan José Fallas-Monge
Escuela de Matemática, Instituto Tecnológico de Costa Rica, Cartago, Costa Rica
2 shared publications Anatoli Torokhti
1 shared publication
Latest Publications
- Open Access
- 7 April 2026
- Published by Instituto Tecnologico de Costa Rica in Revista Tecnología en Marcha
- p. ág. 60-69
Ciudades a la Vista: UAVs Autónomos para Mapas 3D sin LiDAR
En Costa Rica, la topografía urbana enfrenta el desafío de contar con sistemas LiDAR costosos y poco confiables en condiciones adversas como niebla o lluvia. Para superar estas limitaciones, se desarrolló una plataforma de UAVs autónomos de bajo costo que, equipados únicamente con cámaras RGB e IMU, generan mapas tridimensionales urbanos. La propuesta integra técnicas híbridas de estimación monocular de profundidad, combinando aprendizaje auto-supervisado y transferencia de conocimiento, junto con algoritmos de exploración colaborativa basados en enjambres, curvas de Bézier y modelado de feromonas en Neo4J. Tras validar el diseño en simulaciones (PyBullet/Pygame) y vuelos reales interiores, el sistema alcanzó una precisión de decenas de centímetros en profundidad, produjo nubes de puntos georreferenciadas y permitió segmentar semánticamente tráfico y obstáculos con ViT y YOLOv8. Los resultados demuestran que esta aproximación ofrece una alternativa viable y económica al LiDAR tradicional, con potencial para desplegar enjambres reales y optimizar recursos.
- 7 April 2026
- Published by Instituto Tecnologico de Costa Rica in Revista Tecnología en Marcha
- p. ág. 60-69
Open AccessCiudades a la Vista: UAVs Autónomos para Mapas 3D sin LiDAR
En Costa Rica, la topografía urbana enfrenta el desafío de contar con sistemas LiDAR costosos y poco confiables en condiciones adversas como niebla o lluvia. Para superar estas limitaciones, se desarrolló una plataforma de UAVs autónomos de bajo costo que, equipados únicamente con cámaras RGB e IMU, generan mapas tridimensionales urbanos. La propuesta integra técnicas híbridas de estimación monocular de profundidad, combinando aprendizaje auto-supervisado y transferencia de conocimiento, junto con algoritmos de exploración colaborativa basados en enjambres, curvas de Bézier y modelado de feromonas en Neo4J. Tras validar el diseño en simulaciones (PyBullet/Pygame) y vuelos reales interiores, el sistema alcanzó una precisión de decenas de centímetros en profundidad, produjo nubes de puntos georreferenciadas y permitió segmentar semánticamente tráfico y obstáculos con ViT y YOLOv8. Los resultados demuestran que esta aproximación ofrece una alternativa viable y económica al LiDAR tradicional, con potencial para desplegar enjambres reales y optimizar recursos.
- Open Access
- 6 March 2026
- Published by Instituto Tecnologico de Costa Rica in Revista Tecnología en Marcha
- p. ág. 310-324
Proceso de percepción de los alrededores de sistemas UAV en entornos urbanos
Los sistemas de vehículos aéreos no tripulados (UAV) han ganado importancia en diversas aplicaciones debido a su capacidad para operar en entornos peligrosos o de difícil acceso. Sin embargo, la navegación segura y eficiente en entornos complejos como áreas urbanas o interiores de edificios plantea desafíos significativos en términos de percepción del entorno. La percepción precisa del entorno circundante es fundamental para que un UAV pueda construir un mapa 3D, detectar y evitar obstáculos, localizar puntos de interés y planificar trayectorias de vuelo seguras. Para abordar estos desafíos, los sistemas UAV integran una variedad de sensores, como cámaras RGB y de profundidad, radares, lídares y sensores in-erciales. Estos sensores permiten capturar información visual, geométrica y de movimiento del entorno. Además, se emplean técnicas avanzadas de procesamiento de señales, visión por computadora, aprendizaje automático y fusión sensorial para interpretar los datos de los sensores y construir una representación precisa del entorno en 3D. En el trabajo presentado, se logró generar una propuesta para la exploración de caminos en el interior de un espacio mediante la detección de puertas y pasillos utilizando YOLO, en conjunto con la estimación de fondo monocular. Se validaron los caminos y, con esta información, se generaron mapas para explorar un territorio. Se obtuvo una tasa de acierto del 86,9%, con algunos falsos positivos donde el modelo no detectó correctamente la profundidad. El mapa generado guarda un historial del agente que ha explorado el territorio, sirviendo como una opción para integrar con algoritmos de enjambre.
- 6 March 2026
- Published by Instituto Tecnologico de Costa Rica in Revista Tecnología en Marcha
- p. ág. 310-324
Open AccessProceso de percepción de los alrededores de sistemas UAV en entornos urbanos
Los sistemas de vehículos aéreos no tripulados (UAV) han ganado importancia en diversas aplicaciones debido a su capacidad para operar en entornos peligrosos o de difícil acceso. Sin embargo, la navegación segura y eficiente en entornos complejos como áreas urbanas o interiores de edificios plantea desafíos significativos en términos de percepción del entorno. La percepción precisa del entorno circundante es fundamental para que un UAV pueda construir un mapa 3D, detectar y evitar obstáculos, localizar puntos de interés y planificar trayectorias de vuelo seguras. Para abordar estos desafíos, los sistemas UAV integran una variedad de sensores, como cámaras RGB y de profundidad, radares, lídares y sensores in-erciales. Estos sensores permiten capturar información visual, geométrica y de movimiento del entorno. Además, se emplean técnicas avanzadas de procesamiento de señales, visión por computadora, aprendizaje automático y fusión sensorial para interpretar los datos de los sensores y construir una representación precisa del entorno en 3D. En el trabajo presentado, se logró generar una propuesta para la exploración de caminos en el interior de un espacio mediante la detección de puertas y pasillos utilizando YOLO, en conjunto con la estimación de fondo monocular. Se validaron los caminos y, con esta información, se generaron mapas para explorar un territorio. Se obtuvo una tasa de acierto del 86,9%, con algunos falsos positivos donde el modelo no detectó correctamente la profundidad. El mapa generado guarda un historial del agente que ha explorado el territorio, sirviendo como una opción para integrar con algoritmos de enjambre.
- Open Access
- 5 November 2025
- Published by Elsevier in Journal of Computational Mathematics and Data Science
- Vol. 17
A proximal Gauss–Seidel algorithm for solving a generalized low-tubal-rank tensor approximation problem based on the t -product
- 5 November 2025
- Published by Elsevier in Journal of Computational Mathematics and Data Science
- Vol. 17
Open AccessA proximal Gauss–Seidel algorithm for solving a generalized low-tubal-rank tensor approximation problem based on the t -product
- Open Access
- 11 June 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (12) , 1945
Decrease in Computational Load and Increase in Accuracy for Filtering of Random Signals
This paper describes methods for optimal filtering of random signals that involve large matrices. We developed a procedure that allows us to significantly decrease the computational load associated with numerically implementing the associated filter and increase its accuracy. The procedure is based on the reduction of a large covariance matrix to a collection of smaller matrices. This is done in such a way that the filter equation with large matrices is equivalently represented by a set of equations with smaller matrices. The filter we developed is represented by and minimizes the associated error over all matrices . As a result, the proposed optimal filter has two degrees of freedom that increase its accuracy. They are associated, first, with the optimal determination of matrices and second, with an increase in the number p of components in the filter. The error analysis and results of numerical simulations are provided.- 11 June 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (12) , 1945
Open AccessDecrease in Computational Load and Increase in Accuracy for Filtering of Random Signals
This paper describes methods for optimal filtering of random signals that involve large matrices. We developed a procedure that allows us to significantly decrease the computational load associated with numerically implementing the associated filter and increase its accuracy. The procedure is based on the reduction of a large covariance matrix to a collection of smaller matrices. This is done in such a way that the filter equation with large matrices is equivalently represented by a set of equations with smaller matrices. The filter we developed is represented by and minimizes the associated error over all matrices . As a result, the proposed optimal filter has two degrees of freedom that increase its accuracy. They are associated, first, with the optimal determination of matrices and second, with an increase in the number p of components in the filter. The error analysis and results of numerical simulations are provided. - Open Access
- 21 March 2025
- Published by MDPI AG in Preprints.org
Decrease in Computational Load and Increase in Accuracy for Filtering of Random Signals
This paper describes methods for optimal filtering of random signals that involve large matrices. We develop a procedure that allows us to significantly decrease the computational load needed to numerically realize the associated filter and increase the associated accuracy. The procedure is based on the reduction of a large covariance matrix to a collection of smaller matrices. It is done in such a way that the filter equation with large matrices is equivalently represented by a set of equations with smaller matrices. The filter Fp we develop is represented by Fp(v1,…vp)=∑j=1pMjvj and minimizes the associated error over all matrices M1,…,Mp. As a result, the proposed optimal filter has two degrees of freedom to increase the associated accuracy. They are associated, first, with the optimal determination of matrices M1,…,Mp and second, with the increase in the number p of components in the filter Fp. The error analysis and results of numerical simulations are provided.
- 21 March 2025
- Published by MDPI AG in Preprints.org
Open AccessDecrease in Computational Load and Increase in Accuracy for Filtering of Random Signals
This paper describes methods for optimal filtering of random signals that involve large matrices. We develop a procedure that allows us to significantly decrease the computational load needed to numerically realize the associated filter and increase the associated accuracy. The procedure is based on the reduction of a large covariance matrix to a collection of smaller matrices. It is done in such a way that the filter equation with large matrices is equivalently represented by a set of equations with smaller matrices. The filter Fp we develop is represented by Fp(v1,…vp)=∑j=1pMjvj and minimizes the associated error over all matrices M1,…,Mp. As a result, the proposed optimal filter has two degrees of freedom to increase the associated accuracy. They are associated, first, with the optimal determination of matrices M1,…,Mp and second, with the increase in the number p of components in the filter Fp. The error analysis and results of numerical simulations are provided.
- Open Access
- 7 March 2025
- Published by MDPI AG in Mathematical and Computational Applications
- Vol. 30 (2) , 26
Modeling of Nonlinear Systems: Method of Optimal Injections
In this paper, a nonlinear system is interpreted as an operator transforming random vectors. It is assumed that the operator is unknown and the random vectors are available. It is required to find a model of the system represented by a best constructive operator approximation. While the theory of operator approximation with any given accuracy has been well elaborated, the theory of best constrained constructive operator approximation is not so well developed. Despite increasing demands from various applications, this subject is minimally tractable because of intrinsic difficulties with associated approximation techniques. This paper concerns the best constrained approximation of a nonlinear operator in probability spaces. The main conceptual novelty of the proposed approach is that, unlike the known techniques, it targets a constructive optimal determination of all ingredients of the approximating operator where p is a nonnegative integer. The solution to the associated problem is represented by a combination of new best approximation techniques with a special iterative procedure. The proposed approximating model of the system has several degrees of freedom to minimize the associated error. In particular, one of the specific features of the developed approximating technique is special random vectors called injections. It is shown that the desired injection is determined from the solution of a special Fredholm integral equation of the second kind. Its solution is called the optimal injection. The determination of optimal injections in this way allows us to further minimize the associated error.- 7 March 2025
- Published by MDPI AG in Mathematical and Computational Applications
- Vol. 30 (2) , 26
Open AccessModeling of Nonlinear Systems: Method of Optimal Injections
In this paper, a nonlinear system is interpreted as an operator transforming random vectors. It is assumed that the operator is unknown and the random vectors are available. It is required to find a model of the system represented by a best constructive operator approximation. While the theory of operator approximation with any given accuracy has been well elaborated, the theory of best constrained constructive operator approximation is not so well developed. Despite increasing demands from various applications, this subject is minimally tractable because of intrinsic difficulties with associated approximation techniques. This paper concerns the best constrained approximation of a nonlinear operator in probability spaces. The main conceptual novelty of the proposed approach is that, unlike the known techniques, it targets a constructive optimal determination of all ingredients of the approximating operator where p is a nonnegative integer. The solution to the associated problem is represented by a combination of new best approximation techniques with a special iterative procedure. The proposed approximating model of the system has several degrees of freedom to minimize the associated error. In particular, one of the specific features of the developed approximating technique is special random vectors called injections. It is shown that the desired injection is determined from the solution of a special Fredholm integral equation of the second kind. Its solution is called the optimal injection. The determination of optimal injections in this way allows us to further minimize the associated error. - Open Access
- 1 January 2025
- Published by Wiley in Computational and Mathematical Methods
- Vol. 2025 (1)
C-Product Toolbox : A Computational Package for Third‐Order Tensor Operations Based on the Reduced c ‐Product
This paper introduces the C-Product Toolbox , a new computational package available for MATLAB and Python, designed to perform operations on third‐order tensors using a tensor product known as the reduced c ‐product. The reduced c ‐product is a variant of the known c ‐product, a tensor product based on the discrete cosine transform and belonging to a family of tensor products defined by invertible linear transformations. This work presents the theoretical development of the reduced c ‐product used and provides a detailed explanation of each tensor operation implemented in the computational package. Additionally, numerical experiments compare the C-Product Toolbox with an existing MATLAB toolbox that employs the t ‐product, a tensor product based on the discrete Fourier transform. The numerical experiments in this paper demonstrate that the C-Product Toolbox offers superior computational efficiency, achieving faster execution times and lower memory consumption. Furthermore, the practical advantages of the proposed methods are highlighted through an application in video denoising, showcasing the effectiveness of the toolbox in real‐world scenarios.
- 1 January 2025
- Published by Wiley in Computational and Mathematical Methods
- Vol. 2025 (1)
Open AccessC-Product Toolbox : A Computational Package for Third‐Order Tensor Operations Based on the Reduced c ‐Product
This paper introduces the C-Product Toolbox , a new computational package available for MATLAB and Python, designed to perform operations on third‐order tensors using a tensor product known as the reduced c ‐product. The reduced c ‐product is a variant of the known c ‐product, a tensor product based on the discrete cosine transform and belonging to a family of tensor products defined by invertible linear transformations. This work presents the theoretical development of the reduced c ‐product used and provides a detailed explanation of each tensor operation implemented in the computational package. Additionally, numerical experiments compare the C-Product Toolbox with an existing MATLAB toolbox that employs the t ‐product, a tensor product based on the discrete Fourier transform. The numerical experiments in this paper demonstrate that the C-Product Toolbox offers superior computational efficiency, achieving faster execution times and lower memory consumption. Furthermore, the practical advantages of the proposed methods are highlighted through an application in video denoising, showcasing the effectiveness of the toolbox in real‐world scenarios.
- 13 September 2024
- Published by International Scientific Research Publications in Journal of Mathematics and Computer Science
- Vol. 37 (01) , 59-81
A fast method to estimate the Moore-Penrose inverse for well-determined numerical rank matrices based on the Tikhonov regularization
This paper introduces a novel approach for estimating the Moore-Penrose inverse. The method proposed relies on Tikhonov regularization, which requires the computation of all positive singular values of an \(m \times n\) matrix. Additionally, we present a highly efficient and accurate procedure for estimating these singular values. This procedure assumes the well-determined numerical rank of matrices \(A^*A\) (if \(m \geq n\)) and \(AA^*\) (if \(m \leq n\)). Furthermore, we demonstrate the application of our proposed method in solving linear discrete well-posed problems. The paper concludes with numerical simulations to illustrate the advantages of our novel approach. Notably, we compare the execution time associated with our technique to that of some relevant methods in the existing literature, demonstrating that our method outperforms others in terms of computational efficiency. To further substantiate our findings, we conduct computational experiments to measure execution time and speedup. The results affirm the efficiency of our proposed method, showcasing reduced execution times compared to other methods. This contributes to establishing our approach's practical viability and effectiveness in diverse applications.- 13 September 2024
- Published by International Scientific Research Publications in Journal of Mathematics and Computer Science
- Vol. 37 (01) , 59-81
A fast method to estimate the Moore-Penrose inverse for well-determined numerical rank matrices based on the Tikhonov regularization
This paper introduces a novel approach for estimating the Moore-Penrose inverse. The method proposed relies on Tikhonov regularization, which requires the computation of all positive singular values of an \(m \times n\) matrix. Additionally, we present a highly efficient and accurate procedure for estimating these singular values. This procedure assumes the well-determined numerical rank of matrices \(A^*A\) (if \(m \geq n\)) and \(AA^*\) (if \(m \leq n\)). Furthermore, we demonstrate the application of our proposed method in solving linear discrete well-posed problems. The paper concludes with numerical simulations to illustrate the advantages of our novel approach. Notably, we compare the execution time associated with our technique to that of some relevant methods in the existing literature, demonstrating that our method outperforms others in terms of computational efficiency. To further substantiate our findings, we conduct computational experiments to measure execution time and speedup. The results affirm the efficiency of our proposed method, showcasing reduced execution times compared to other methods. This contributes to establishing our approach's practical viability and effectiveness in diverse applications.- 18 April 2024
- Published by Elsevier in Journal of Computational and Applied Mathematics
- Vol. 448
Fast random vector transforms in terms of pseudo-inverse within the Wiener filtering paradigm
- 18 April 2024
- Published by Elsevier in Journal of Computational and Applied Mathematics
- Vol. 448
Fast random vector transforms in terms of pseudo-inverse within the Wiener filtering paradigm
- Open Access
- 21 January 2024
- Published by Universidad Catolica del Norte - Chile in Proyecciones (Antofagasta)
- Vol. 43 (1) , 189-224
Optimal modeling of nonlinear systems
Our work addresses a development and justification of the new approach to the modeling of nonlinear systems. Let $\f$ be an unknown input-output map of the system with a random input and output $\y$ and $\x$, respectively. It is assumed that $\y$ and $\x$ are available and covariance matrices formed from $\y$ and $\x$ are known. We determine a model of $\f$ so that an associated error is minimized. To this end, the model $\ttt_p$ is constructed as a sum of $p+1$ particular parts, in the form $\ttt_p (\y) = \sum_{j=0}^{p}G_j H_j Q_j(\vv_j)$ where $G_j$ and $ H_j$, for $j=0,\ldots, p$, are matrices to be determined, and $\vv_j$, for $j=1,\ldots,p$, is a special random vector called the injection. We denote $\vv_0=\y$. Injections $\vv_1,\ldots, \vv_p$ are aimed to diminish the error associated with the proposed model $\ttt_p$. Further, $Q_j$ is a special transform aimed to facilitate the numerical realization of model $\ttt_p$. It is determined in the way allowing us to optimally determine $G_j$ and $ H_j$ as a solution of $p+1$ separate error minimization problems which are simpler than the original minimization problem. The empirical determination of injections $\vv_1,\ldots, \vv_p$ is considered. The proposed method has several degrees of freedom to diminish the associated error. They are `degree' $p$ of $\ttt_p$, choice of matrices $G_0, H_0, \ldots,$ $G_p, H_p$, dimensions of matrices $G_0, H_0, \ldots,$ $G_p, H_p$ and injections $\vv_1,\ldots, \vv_p$, respectively. Four numerical examples are provided. At the end, the open problem is formulated.- 21 January 2024
- Published by Universidad Catolica del Norte - Chile in Proyecciones (Antofagasta)
- Vol. 43 (1) , 189-224
Open AccessOptimal modeling of nonlinear systems
Our work addresses a development and justification of the new approach to the modeling of nonlinear systems. Let $\f$ be an unknown input-output map of the system with a random input and output $\y$ and $\x$, respectively. It is assumed that $\y$ and $\x$ are available and covariance matrices formed from $\y$ and $\x$ are known. We determine a model of $\f$ so that an associated error is minimized. To this end, the model $\ttt_p$ is constructed as a sum of $p+1$ particular parts, in the form $\ttt_p (\y) = \sum_{j=0}^{p}G_j H_j Q_j(\vv_j)$ where $G_j$ and $ H_j$, for $j=0,\ldots, p$, are matrices to be determined, and $\vv_j$, for $j=1,\ldots,p$, is a special random vector called the injection. We denote $\vv_0=\y$. Injections $\vv_1,\ldots, \vv_p$ are aimed to diminish the error associated with the proposed model $\ttt_p$. Further, $Q_j$ is a special transform aimed to facilitate the numerical realization of model $\ttt_p$. It is determined in the way allowing us to optimally determine $G_j$ and $ H_j$ as a solution of $p+1$ separate error minimization problems which are simpler than the original minimization problem. The empirical determination of injections $\vv_1,\ldots, \vv_p$ is considered. The proposed method has several degrees of freedom to diminish the associated error. They are `degree' $p$ of $\ttt_p$, choice of matrices $G_0, H_0, \ldots,$ $G_p, H_p$, dimensions of matrices $G_0, H_0, \ldots,$ $G_p, H_p$ and injections $\vv_1,\ldots, \vv_p$, respectively. Four numerical examples are provided. At the end, the open problem is formulated.