Jose Maria Manzano
@uloyola.es
Lecturer - Engineering Department
Universidad Loyola
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Beatrice Sonzogni, José María Manzano, Marco Polver, Fabio Previdi, and Antonio Ferramosca
Elsevier BV
J. M. Nadales, J. M. Manzano, A. Barriga, and D. Limon
Institute of Electrical and Electronics Engineers (IEEE)
One of the main open challenges in the field of learning-based control is the design of computing architectures able to process data in an efficient way. This is of particular importance when time constraints must be met, as, for instance, in real-time decision-making systems operating at high frequencies or when a vast amount of data must be processed. In this respect, field-programmable gate array (FPGA)-based parallel processing architectures have been hailed as a potential solution to this problem. In this article, a low-level design methodology for the implementation on FPGA platforms of Lipschitz interpolation (LI) algorithms is presented. The proposed design procedure exploits the potential parallelism of the LI algorithm and allows the user to optimize the area and energy resources of the resulting implementation. Besides, the proposed design allows to know in advance a tight bound of the error committed by the FPGA due to the representation format. Therefore, the resulting implementation is a highly parallelized and a fast architecture with an optimal use of the resources and consumption and with a fixed numerical error bound. These facts flawlessly suit the desirable specifications of learning-based control devices. As an illustrative case study, the proposed algorithm and architecture have been used to learn a nonlinear model predictive control law applied to self-balance a two-wheel robot. The results show how computational times are several orders of magnitude reduced by employing the proposed parallel architecture, rather than sequentially running the algorithm on an embedded ARM-CPU-based platform.
Guillermo Bejarano, José María Manzano, José Ramón Salvador, and Daniel Limon
Elsevier BV
J.M. Manzano, J.R. Salvador, J.B. Romaine, and L. Alvarado-Barrios
Elsevier BV
J.M. Manzano, Luis Orihuela, Erid Pacheco, and Mario Pereira
Elsevier BV
J.M. Manzano, J. Bareiro, G.B. Cáceres, J.R. Salvador, and P. Millán
Elsevier BV
Thalia A. Morel, Guillermo Bejarano, Jose Maria Manzano, and Luis Orihuela
IEEE
This paper describes the application and experimental validation of two different extended state observers for autonomous surface vehicles. Only the vessel position and orientation are assumed to be measured, being both estimators able to recover, not only the velocities, but also the so-called lumped generalised disturbance, which groups both environmental disturbances and nonlinear/unmodelled vessel dynamics. The experimental platform and the open-loop experiments performed to assess and validate the performance of both estimators are described. The experimental results confirm the prospects given by their previously published simulation results, though some room for improvement is detected in the estimation of the state dynamics.
Jose Maria Manzano, David Muñoz de la Peña, and Daniel Limon
Wiley
J. M. Manzano, D. Muñoz de la Peña, J. Calliess, and D. Limon
Wiley
A data‐based predictive controller is proposed, offering both robust stability guarantees and online learning capabilities. To merge these two properties in a single controller, a double‐prediction approach is taken. On the one hand, a safe prediction is computed using Lipschitz interpolation on the basis of an offline identification dataset, which guarantees safety of the controlled system. On the other hand, the controller also benefits from the use of a second online learning‐based prediction as measurements incrementally become available over time. Sufficient conditions for robust stability and constraint satisfaction are given. Illustrations of the approach are provided in a simulated case study.
Jose Maria Manzano, David Munoz de la Pena, Jan-Peter Calliess, and Daniel Limon
Institute of Electrical and Electronics Engineers (IEEE)
This article presents a novel learning method based on componentwise Hölder continuity, which allows one to consider independently the contribution of each input to each output of the function to be learned. The method provides a bounded prediction error, and its learning property is proven. It can be used to obtain a predictor for a nonlinear robust learning-based predictive controller for constrained systems. The resulting controller achieves better closed loop performance and larger domains of attraction than learning methods that only consider nonlinear set membership, as illustrated by a case study.
José María Manzano, David Muñoz de la Peña, and Daniel Limon
Elsevier BV
Jose Maria Manzano, Jose Ramon Salvador, Guillermo Bejarano, and Daniel Limon
IEEE
This paper proposes a nonlinear model predictive control-based guidance strategy for autonomous surface vehicles, focused on the path following approach to motion control. The application of this strategy, in addition to overcome the drawbacks of previous line-of-sight-based guidance laws, intends to enable the application of predictive strategies also to the low-level control, responsible for tracking the references provided by the guidance high-level strategy. The robust and stable features of the proposed strategy are discussed, while the effectiveness and the advantages over other nonlinear guidance laws are illustrated through a complete set of simulations.
J.M. Nadales, J.M. Manzano, A. Barriga, and D. Limon
IEEE
The implementation of nonlinear model predictive controllers for systems operating at high frequencies constitutes a significant challenge, mainly because of the complexity and time consumption of the optimization problem involved. An alternative that has been proposed is the employment of data-driven techniques to offline learn the control law, and then to implement it on a target embedded platform. Following this trend, in this paper we propose the implementation of predictive controllers on FPGA platforms making use of a parallel version of the machine learning technique known as Lipschitz interpolation. By doing this, computation time can be enormously accelerated. The results are compared to those obtained when the sequential algorithm runs on standard CPU platforms, and when the system is controlled by solving the optimization problem online, in terms of the error made and computing time. This method is validated in a case study where the nonlinear model predictive controller is employed to control a self-balancing two-wheel robot.
José María Manzano, Daniel Limon, David Muñoz de la Peña, and Jan-Peter Calliess
Elsevier BV
J.M. Manzano, J. Calliess, D. Muñoz de la Peña, and D. Limon
Elsevier BV
J.M. Manzano, D. Limon, D. Munoz de la Pena, and J.P. Calliess
IEEE
The authors have recently developed predictive controllers based on prediction models derived from experimental data, by means of a class of Hölder interpolation called kinky inference. This paper provides a step forward by proposing a novel estimation method based on componentwise Hölder interpolation. This allows to explicitly consider the contribution of each component on each output, yielding better estimations. Following the procedure used in previous works, this estimation method is used to provide a predictor for a nonlinear robust data-based predictive controller, whose performance and robustness is enhanced by the new setting. The properties of the proposed controller are demonstrated in a case study.
J.M. Manzano, J.M. Nadales, D. Munoz de la Pena, and D. Limon
IEEE
This paper deals with an economic predictive controller for the optimal operation of a plant under the assumption that the only measurement of the system is the economic cost function to be minimized. In order to predict the evolution of the economic cost for a given input trajectory, an oracle with a NARX structure is proposed. Sufficient conditions to ensure the existence of such oracle are given, and based on this oracle, a novel predictive controller is proposed. Under certain assumptions, including ideal accurate estimation, it is proven that the proposed oracle-based economic predictive controller provides the same solution of a standard economic MPC based on the model plant, inheriting the properties of this class of controllers. The proposed oracle-based economic predictive controller is applied to a quadruple-tank process example.
A. Blaas, J.M. Manzano, D. Limon, and J. Calliess
IEEE
Their flexibility to learn general function classes renders nonparametric regression algorithms particularly attractive in system identification and data-based control settings, where little a priori knowledge about a dynamical system is to be presumed. Building on approaches known as NSM-or Lipschitz regression, we propose a new nonparametic machine learning approach. While it inherits theoretical learning guarantees from the methods it is built upon, it is designed to limit the computational effort both for learning and for generating predictions. This renders our method applicable to online system identification and control settings where the desired sample frequency precludes previous nonparametric approaches from being deployed. Apart from deriving a guarantee on the ability of our method to learn any continuous function, we illustrate some of its practical merits on a number of benchmark comparison problems.
José María Manzano, Daniel Limón, David Muñoz de la Peña, and Jan Peter Calliess
Institution of Engineering and Technology (IET)
In this study, the authors propose a stabilising data-based model predictive controller for systems subject to constraints in which the prediction model is inferred from experimental data of the plant using a machine learning technique. The inference method is a modification of the kinky inference tailored for model predictive control. In particular, the modified method has a lower computational effort and provides smoother predictions than the original method. The controller formulation considers soft constraints in the outputs, hard constraints in the inputs and guarantees closed-loop robust stability as well as performance by means of the use of different control and prediction horizons and a weighted terminal cost. Under the assumption that the model of the system is Holder continuous, they prove that the closed-loop system is input-to-state stable with respect to the estimation errors. The results are demonstrated in a case study of a continuously stirred-tank reactor.
T. Alamo, J. M. Manzano, and E. F. Camacho
Springer International Publishing
J.M. Manzano, D. Limon, D. Muñz de la Peñ, and J. Calliess
Elsevier BV
Michael Maiworm, Daniel Limon, Jose Maria Manzano, and Rolf Findeisen
Elsevier BV