Search

International audience

International audience; This paper addresses the implementation of a novel approach to the design of Model Predictive Control (MPC) on a brandy distillation process. This new approach is based on multi-objective optimization with prioritization and switching between multiple models obtained at different operating phases. At each sampling step, several conflicting objectives are to be optimized under an order of priority that may be changed from one phase of the process working cycle to another. The proposed Switched Multi-Objective Prioritizing MPC (S-MOP-MPC) is compared with a PID-based control structure in terms of various performance indices. The results show that S-MOP-MPC provides better control and better performance than the PID-based control structure.

Summary In this paper, we provide a general framework for robust optimal estimation over a lossy and delayed network. A threshold principle is introduced to integrate network-induced uncertainties into packet losses, which are modeled with a Bernoulli process. Based on stability conditions derived from two Riccati equations, we show the existence of critical observation arrival probabilities below which the optimal estimator stochastically fails to converge. Moreover, the result is extended to a real system with variable process disturbance, which has an indicator for its admissible bound in terms of a given restriction of estimation accuracy. The proposed method is experimented on a specific automobile application, the battery state of charge estimation. Copyright © 2015 John Wiley & Sons, Ltd.

International audience; This paper investigates H∞ predictive control problem for a class of networked control system, and presents a method for quantitatively selecting the control weighting parameter. Based on Lyapunov function approach and H∞ control theory, a generalized predictive state feedback controller is derived to guarantee the closed-loop networked system stable. Then, an algebraic method for selecting a feasible GPC control-weighting parameter is proposed. The contribution realizes the adaptive parameter tuning for GPC with respect to response demands and performance. A simulation example demonstrates the feasibility and practicability of the proposed method.

International audience; This article deals with a method for providing the best estimate as possible of a nonlinear system state. It is concerned with both finding the most accurate state estimator and obtaining the best measurements as possible, thanks to the redundancy of sensors under hardware and economical constraints. These choices are algorithmically done. The selection of the state estimator takes place according to the variances of estimation errors. Concerning the elaboration of hardware redundancy, it is done by computing a new criterion evaluating estimation quality with respect to sensors losses. The proposed algorithm is then applied on the Lorenz system, parameterized to present a chaotic behaviour, which constitutes a priori difficult situation.

Optimal energy management in HEV is primordial. A key element of energy management strategies is the battery pack State-Of-Charge (SOC). In a real-time setting, a most commonly used tool to estimate the SOC is the Kalman filter-based techniques, and more especially adaptive Kalman filtering. In this paper, we consider a particular class of Linear Time-Invariant (LTI) Single-Input Single-Output (SISO) system suited to a lot of battery types. Then, we propose a new form of adaptive Kalman filter which is more convenient to use than the original adaptive Kalman filter. Finally, we highlight the effectiveness of the proposed approach on the SOC estimation problem of a four lead-acid battery pack.

This paper addresses the problem of noise tolerance of a linear estimator for discrete-time networked control systems with incomplete observations and process noise with variable intensity (here expressed by its covariance). Through modified Riccati equation with packet arrival rate and linear matrix inequality theorem (LMI), our approach is to first provide an analytic relation between the filter convergence and the packet arrival process. Then a robust bound of packet arrival rate is proved to exist. We design a dropping over-delayed packet strategy so as to satisfy robust packet arrival rate. Moreover, unified algorithms to find the tradeoff with the process noise with admissible intensity is provided in terms of a given restriction of admitted state estimation error-covariance. Such a tool can help in modern industrial networked control applications. Examples demonstrate its effectiveness.

International audience; In this paper, with regards to discrete-time networked control systems with incomplete observations, a novel optimal linear estimator design is presented. Through a threshold-length buffer coupled to the estimator and a dropping over-delayed packet strategy, the packet arrival process is artificially limited to time pre-threshold at current moment. According to statistics and probability theory, we model the packet arrival probability subject to the choice of threshold. Based on linear matrix inequality (LMI) theorem and modified Riccati equation with packet arrival probability, with the existence of a critical bound of packet arrival rate, we can completely establish the relation between the filter convergence and the threshold-based packet arrival process as well as characterize the filter performance. Examples demonstrate the feasibility and practicability of the proposed method.

International audience; With regards to the characteristics of uncertain long delay of networked control system (NCS), the stochastic time- delay system can be transformed to a deterministic delay system by placing proper quantities of buffers at the nodes of the double-end networked control systems. An observer-based control design subject to fixed time delay is proposed to construct the augmented state-space model. Based on the Lyapunov function and linear matrix inequality (LMI) theorem, sufficient and necessary conditions for stability of such systems using static state feedback are given. Simulation result proves its effectiveness.

International audience; This paper addresses the problem of designing a classical observer extended to dynamical discrete-time systems with constant delays. The idea of two forms of state observers is presented to compensate fixed time-delay. Convergence conditions have been found to ensure the stability of the error dynamics and the related gains may be selected based on a set of linear matrix inequality (LMI). The results show that the compensation-based state observer is much superior in dealing with time-delay in the case that the buffer permits us to transform the time-delay system into a fixed time-delay system. Simulation and experimental results are provided to demonstrate the advantage of the proposed observation technique.

This paper presents a method for dealing with Kalman filtering under particular classes of nonlinear state model and state inequality constraints. This method is based on a modification of the optimal Kalman gain in order to enforce the state constraints if necessary. The proposed method is applied to a nonlinear hydraulic system in order to highlight the efficiency of this method. Copyright © 2010 John Wiley & Sons, Ltd.

International audience; Cet article traite du problème d'estimation d'état en présence d'entrées inconnues. Actuellement, il existe différentes approches pour aborder un tel problème, mais étant soit prises en défaut en cas de pertes de mesures, soit applicables à une classe restreinte de systèmes. Le but de cet article est de proposer une approche d'estimation d'état en présence d'entrées inconnues basée sur une modélisation de ces dernières. Le modèle développé est de type dynamique et est conçu de manière à ce que les entrées inconnues présentent une variance d'Allan plate. L'approche proposée a été expérimentée avec succès sur une application de navigation hybridée inertie-GPS.

International audience; This paper proposes a general method for dealing with state estimation under nonlinear state soft inequality constraints. This method is based on the projection approach, and then has the advantage to be compatible with any kind of state estimator. In order to be taken into account, the nonlinear constraints are linearized about the constrained state using an iterated approach. The proposed algorithm is tested on a three-dimension tracking application with nonlinear constraints on the moving body acceleration. The results are compared with those of an unconstrained Kalman filter.