Demenkov M.,Institute of Control science
Proceedings of 1st International Conference on Event-Based Control, Communication and Signal Processing, EBCCSP 2015 | Year: 2015
A hot topic of research in the model predictive control community is the real-Time implementation of MPC controllers for 'fast' systems (e.g. in automotive and aerospace applications). The controller needs to optimize a control sequence on-line while maintaining strict constraints on the time of computations. In this work-in-progress paper we discuss the possibility of asynchronous MPC computations while maintaining the stability of the closed-loop system. In case the computations cannot be performed within the available time slot and this can affect the stability of the closed-loop system, a second controller is activated, which is based on the explicitly constructed MPC cost function. In the case of continuous linear plant dynamics and polyhedral cost, this approach leads to the easily implementable relay-Type control. © 2015 IEEE.
Novikov D.A.,Institute of Control science
Advances in Intelligent Systems and Computing | Year: 2016
Historical and modern approaches towards heterogeneous models and hierarchical modeling in control theory are listed. A unified approach towards a design of hierarchical models of multiagent systems is described and illustrated by the example of diffuse bomb model. Problems of “hierarchical automation” in organization-technical and production systems are described. General problems of hierarchical modeling are underlined. © Springer International Publishing Switzerland 2016.
Khlebnikov M.V.,Institute of Control Science
IFAC-PapersOnLine | Year: 2015
In this talk, the problem of robust stabilization of bilinear control systems is considered. Based on the technique of linear matrix inequalities and quadratic Lyapunov function, the so-called robust stabilizability ellipsoid is designed such that the trajectory of the closed-loop system, starting inside the ellipsoid asymptotically tends to zero for all admissible uncertainties. The results obtained allow to design an approximation of the robust stabilizability domain of the uncertain bilinear control system.
Demenkov M.,Institute of Control science
54th Israel Annual Conference on Aerospace Sciences 2014 | Year: 2014
An active flutter suppression using saturated control is investigated for a 2d of wing section with nonlinear torsional stiffness and saturated control effectors. The suppression of oscillations in the nonlinear closed-loop system is achieved through maximization of the region of attraction for the linearized input-constrained system. For this, we approximate maximum possible region of attraction by a polyhedron and associate with it a level set of the polyhedral Lyapunov function and a corresponding relay control law. Finally, we derive linear saturated control law, which is a relaxation of the relay one. Applying the control to the original nonlinear system, we confirm the local disappearance of oscillations using numerical bifurcation analysis.
Kustov A.,Institute of Control science
Proceedings of 2016 International Conference "Stability and Oscillations of Nonlinear Control Systems" (Pyatnitskiy's Conference), STAB 2016 | Year: 2016
In this paper, the anisotropic norm boundedness sufficient conditions of system with multiplicative noise in input-to-state matrix are presented. It is shown that these conditions can be described by inequality for anisotropic norms of auxiliary systems. The main result is written in the form of linear matrix inequalities and additional algebraic inequalities. © 2016 IEEE.
Salikhov Z.G.,Institute of Control science
IFAC Proceedings Volumes (IFAC-PapersOnline) | Year: 2015
One-criterion problem of control the complex metallurgical process presented by variable parameters in vector notation was considered. In the implementation vector notation presentation transformed into canonical form by means of piecewise continuous functions optimizing control variables on a given interval [0, T]. Pontryagin's maximum principle is used to determine the optimal conditions for solving one-criterion problem, which is formulated as regular multi-criteria (maximum performance, minimum loss of ferrous and nonferrous metals and energy, etc.) optimization problem. Assumed that a piecewise continuous equations in the zone u ⊂ U, and their existence in space U ⊂ E as the main way to solve multi-criteria optimization problems of metallurgical processes is acceptable convolution of trajectory one-criterion optimum problem. In the modeling of singularly perturbed multi-criteria optimal control problems of metallurgical processes was used the method of Whitham or the search of controlled variable and control action carried out in the form of an asymptotic series in small parameter with the highest differential equations derivatives. It is shown that the expansion of control actions equations system in an asymptotic series in the small parameter allows a distributed optimal control problem for considered process to be transformed to recurrent sequence of simple regular tasks. Finding the coefficients of the asymptotic expansion of the optimal control actions for multi-criteria metallurgical process is the desired solution with given initial conditions and singular perturbations. © 2015, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Chervonenkis A.Y.,Institute of Control science
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2011
The problem of reconstructing dependencies from empirical data became very important in a very large range of applications. Procedures used to solve this problem are known as "Methods of Machine Learning" [1,3]. These procedures include methods of regression reconstruction, inverse problems of mathematical physics and statistics, machine learning in pattern recognition (for visual and abstract patterns represented by sets of features) and many others. Many web network control problems also belong to this field. The task is to reconstruct the dependency between input and output data as precisely as possible using empirical data obtained from experiments or statistical observations. © 2011 Springer-Verlag Berlin Heidelberg.
Demenkov M.,Institute of Control science
2015 International Conference on Mechanics - Seventh Polyakhov's Reading | Year: 2015
For a locally stable polynomial dynamical system its region of attraction can be estimated by a polynomial Lyapunov function level set. We formulate this problem in terms of global minimization of a polynomial function over single polynomial constraint. We describe a simple Matlab tool, which employs numerical algebraic geometry methods for computing all local solutions of the optimization problem and therefore its global solution. © 2015 IEEE.
Rapoport L.B.,Institute of Control science
European Control Conference, ECC 1999 - Conference Proceedings | Year: 2015
For mu Iti vari able Lure systems which are not stable in whole the problem of estimating of the attraction domain is considered. For initial conditions which belong to this domain the asymptotic stability 'in large' is guaranteed. This problem has received much attention in the literature (see for example ,  and the references there). The state variable method as well as the frequency domain method were used. In the both cases the problem has been treated by the Lur'e-Postnikov Lyapnnov function with subsequent usage of the S-procednre. For the MTMO case the S-procednre is known to obtain only sufficient conditions for existence of the Lnr'e-Postnikov function. So, the mnltivariable extension of the Popov criterion derived on its base is conservative even in the class of criteria based on the Lnr'e-Postnikov function. We use non-conservative extension of the S-procednre (, ) which makes it possible to derive new absolute stability criterion which is less conservative than mnltivariable Popov criterion based on the usual S-procednre. Using this frequency criterion, the estimation of the attraction domain is constructed. It is less conservative than one obtained in . The LMT approach to verify whether the initial condition belongs to the estimation of the attraction domain is proposed. © 1999 EUCA.
Rusyaeva E.,Institute of Control science
Communications in Computer and Information Science | Year: 2014
We presented sequentially stages of studying the cognitive processes and obtaining scientific knowledge: from traditional social to individual cognitive techniques. Given comparative tables of knowledge levels, subjects' goal tasks during the transition to each level and types of mental entities themselves. Cognitive techniques, "tools" of the cognitive process are stated. © Springer International Publishing Switzerland 2014.