Decomposition of problems of optimal control and estimation for discrete systems with fast and slow variables

Consideration was given to the linear-quadratic problem of optimal control for the discrete linear system with fast and slow variables under incomplete information about system state. Decomposition of the discrete matrix Riccati equations was carried out. The proposed decomposition algorithm relies on a geometrical approach using the properties of the invariant manifolds of slow and fast motions of the nonlinear multirate discrete systems as basis. The splitting transformation was constructed in the form of asymptotic decomposition in the degrees of a small parameter.     

Model reduction of multi-scale chemical Langevin equations

This paper addresses the model reduction problem for a class of stiff chemical Langevin equations that arise as models of biomolecular networks with fast and slow reactions and can be described as continuous Markov processes. Initially, a coordinate transformation is sought that allows the decoupling of fast and slow variables in the model equations. Necessary and sufficient conditions are derived for such a linear transformation to exist, along with an explicit change of variables which achieves the desired decoupling. For the systems for which this step is applicable, the method of adiabatic elimination is applied to determine a representation of the slow dynamics. Theoretical concepts and results are illustrated with simple examples.     

Robust sampled-data H/sub /spl infin// control of linear singularly perturbed systems

State-feedback H/sub /spl infin// control problem for linear singularly perturbed systems with norm-bounded uncertainties is studied. The fast variables are sampled with fast rates, while for the slow variables both cases of slow and of fast sampling are considered. The recent "input delay" approach to sampled-data control is applied, where the closed-loop system is represented as a continuous one with time-varying input delay. Linear matrix inequalities (LMIs) for solution of H/sub /spl infin// control problem are derived via input-output approach to stability and L/sub 2/-gain analysis of time-delay systems. A numerical example illustrates the efficiency of the method.     

Multivariable circle criteria for multiparameter singularlyperturbed systems

A direct approach to the Lur'e problem for singularly perturbed systems (SPS) is proposed. In contrast to previous results, the feedback connection between the linear and nonlinear parts of SPS is allowed to depend essentially on both the slow and the fast variables. The Lur'e problem for multiparameter SPS is studied by the same framework     

Adaptive composite suboptimal control for linear singularly perturbed systems with unknown slow dynamics

In this article, using singular perturbation theory and adaptive dynamic programming (ADP) approach, an adaptive composite suboptimal control method is proposed for linear singularly perturbed systems (SPSs) with unknown slow dynamics. First, the system is decomposed into fast‐ and slow‐subsystems and the original optimal control problem is reduced to two subproblems in different time‐scales. Afterward, the fast subproblem is solved based on the known model of the fast‐subsystem and a fast optimal control law is designed by solving the algebraic Riccati equation corresponding to the fast‐subsystem. Then, the slow subproblem is reformulated by introducing a system transformation for the slow‐subsystem. An online learning algorithm is proposed to design a slow optimal control law by using the information of the original system state in the framework of ADP. As a result, the obtained fast and slow optimal control laws constitute the adaptive composite suboptimal control law for the original SPSs. Furthermore, convergence of the learning algorithm, suboptimality of the adaptive composite suboptimal control law and stability of the whole closed‐loop system are analyzed by singular perturbation theory. Finally, a numerical example is given to show the feasibility and effectiveness of the proposed methods.     

New stability criteria of linear singular systems with time-varying delay

Most of the existing results on the stability problem of delayed singular systems only pertain to the case of constant delay. This is due to the fact that time-varying delay makes it hardly possible to explicitly express the fast variables. In this paper, aiming at dealing with the case of time-varying delay, we create a way to prove the stability by using a perturbation approach. Rather, we first get the decay rate for slow variables by using Lyapunov functional approach and, furthermore, guarantee that the fast variables fall into decay by characterising their effect on the derived decay rate. Also, we present a convexity technique in computing the constructed Lyapunov functional which contributes to the elimination of the possible conservatism caused by the varying rate of delay. Finally, we provide two numerical examples to demonstrate the effectiveness of the method.     

Exact slow–fast decomposition of the nonlinear singularly perturbed optimal control problem

We study the infinite horizon nonlinear quadratic optimal control problem for a singularly perturbed system, which is nonlinear in both, the slow and the fast variables. It is known that the optimal controller for such problem can be designed by finding a special invariant manifold of the corresponding Hamiltonian system. We obtain exact slow–fast decomposition of the Hamiltonian system and of the special invariant manifold into the slow and the fast ones. On the basis of this decomposition we construct high-order asymptotic approximations of the optimal state-feedback and optimal trajectory.     

Linear parameter varying switching attitude control for a near space hypersonic vehicle with parametric uncertainties

This paper deals with the problem of linear parameter varying (LPV) switching attitude control for a near space hypersonic vehicle (NSHV) with parametric uncertainties. First, due to the enormous complexity of the NSHV nonlinear attitude dynamics, a slow–fast loop polytopic LPV attitude model is developed by using Jacobian linearisation and the tensor product model transformation approach. Second, for the purpose of less conservative attitude controller design, the flight envelope is divided into four subregions. For each parameter subregion, slow-loop and fast-loop LPV controllers are designed. By the defined switching character function, these slow–fast loop LPV controllers are then switched in order to guarantee the closed-loop NSHV system to be asymptotically stable and satisfy a specified tracking performance criterion. The condition of LPV switching attitude controller synthesis is given in terms of linear matrix inequalities, which can be readily solved via standard numerical software, and the robust stability analysis of the closed-loop NSHV system is verified based on multiple Lypapunov functions. Finally, numerical simulations have demonstrated the effectiveness of the proposed approach.     

System decomposition technique for spray modelling in CFD codes

A new decomposition technique for a system of ordinary differential equations is suggested, based on the geometrical version of the integral manifold method. This is based on comparing the values of the right hand sides of these equations, leading to the separation of the equations into ‘fast’ and ‘slow’ variables. The hierarchy of the decomposition is allowed to vary with time. Equations for fast variables are solved by a stiff ODE system solver with the slow variables taken at the beginning of the time step. The solution of the equations for the slow variables is presented in a simplified form, assuming linearised variation of these variables for the known time evolution of the fast variables. This can be considered as the first order approximation for the fast manifold. This technique is applied to analyse the explosion of a polydisperse spray of diesel fuel. Clear advantages are demonstrated from the point of view of accuracy and CPU efficiency when compared with the conventional approach widely used in CFD codes. The difference between the solution of the full system of equations and the solution of the decomposed system of equations is shown to be negligibly small for practical applications. It is shown that in some cases the system of fast equations is reduced to a single equation.     

A multirate digital control via a discrete-time observer for non-linear singularly perturbed continuous-time systems

Based on Euler's methodology, we present a new discretization scheme of two-time-scale non-linear continuous-time systems. This scheme of the discretization consists in using two periods, slow and fast, which are determined with respect to slow and fast variables rate respectively, and allows us to define and then implement a multirate digital control for such systems to cope with this multirate measurements on-line. To this end, two reduced order observers are constructed for the slow and fast subsystems by referring to the knowledge of the left-inversion of the state-to-measurements map problem.     

Blue sky catastrophe in systems with nonclassical relaxation oscillations

The feasibility of the well-known blue sky bifurcation in a class of three-dimensional singularly perturbed systems of ordinary differential equations with one fast and two slow variables is studied. A characteristic property of the considered systems is that so-called nonclassical relaxation oscillations occur in them. The same name is used for oscillations with slow components, which are asymptotically close to some time-discontinuous functions and a δ-like fast component. Cases when the blue sky catastrophe results in a stable relaxation cycle or a stable two-dimensional invariant torus are analyzed. The problem of the appearance of homoclinic structures is also considered.     

Robust Analytical Redundancy Relations for Fault Diagnosis In Nonlinear Systems

The problem of robust analytical redundancy relations design for fault diagnosis in nonlinear systems is investigated. An approach is proposed to solve this problem through an extension of the system state‐space by means of constant (or slow varying) unknown parameters, followed by designing analytical redundancy relations that are insensitive both to state variables and unknown parameters. To make the design practicable, some problems related to nonlinear transformation of the extended system model into observable one are considered. The quasi‐linear computational form for robust analytical redundancy relations is described. Fault detectability and distinguishability conditions are formulated, and a recommendation for choosing the size of the moving time window is made. The results are demonstrated through an example involving an SISO nonlinear model of an underwater vehicle.     

A Hybrid Implicit-Explicit Adaptive Multirate Numerical Scheme for Time-Dependent Equations

We develop a hybrid implicit and explicit adaptive multirate time integration method to solve systems of time-dependent equations that present two significantly different scales. We adopt an iteration scheme to decouple the equations with different time scales. At each iteration, we use an implicit Galerkin method with a fast time-step to solve for the fast scale variables and an explicit method with a slow time-step to solve for the slow variables. We derive an error estimator using a posteriori analysis which controls both the iteration number and the adaptive time-step selection. We present several numerical examples demonstrating the efficiency of our scheme and conclude with a stability analysis for a model problem.     

A new reduction technique for a class of singularly perturbedoptimal control problems

We consider a class of singularly perturbed optimal control problems which may not be approximated by the reduced problems constructed via the formal replacement of the fast variables by the states of equilibrium of the “fast” subsystem considered with “frozen” slow variables and controls. We construct a reduced optimal control problem which provides a true approximation for the problems under consideration and write down the necessary and sufficient optimality conditions for this reduced problem     

Distributed Resource Allocation via Accelerated Saddle Point Dynamics

In this paper, accelerated saddle point dynamics is proposed for distributed resource allocation over a multi-agent network, which enables a hyper-exponential convergence rate. Specifically, an inertial fast-slow dynamical system with vanishing damping is introduced, based on which the distributed saddle point algorithm is designed. The dual variables are updated in two time scales, i.e., the fast manifold and the slow manifold. In the fast manifold, the consensus of the Lagrangian multipliers and the tracking of the constraints are pursued by the consensus protocol. In the slow manifold, the updating of the Lagrangian multipliers is accelerated by inertial terms. Hyper-exponential stability is defined to characterize a faster convergence of our proposed algorithm in comparison with conventional primal-dual algorithms for distributed resource allocation. The simulation of the application in the energy dispatch problem verifies the result, which demonstrates the fast convergence of the proposed saddle point dynamics.      

Pre-processing of high-dimensional categorical predictors in classification settings

Models in industrial applications can encounter categorical predictors with a large number of categories (hundreds or thousands). An example is the lot identifier of product in semiconductor manufacturing. Such variables represent a serious problem for practically all modern classification techniques. The goal is an efficient, computationally fast way to discover a small number of natural partitions of values for such variables that have similar statistical properties in terms of categorical response. Such partitions (interesting by itself) can be used then as an input to standard learning algorithms, such as decision trees, support vector machines, etc. The proposed approach introduces a data transformation on derived sparse frequency tables. Application of even simplest non-hierarchical metric clustering method to the transformed coordinates shows significant improvement both in speed and quality of partition in comparison to currently used methods.     

Suboptimal control of singularly perturbed systems and periodicoptimization

A traditional approach to singularly perturbed optimal control problems is based on an approximation of these problems by reduced problems which are obtained via the formal replacement of the fast variables by the states of equilibrium of the fast subsystems considered with frozen slow variables and controls. It is shown that such an approximation is valid if and only if certain families of periodic optimization problems admit steady state solutions. It is also shown how the solutions of these problems can be used to construct suboptimal controls for singularly perturbed problems when approximation by reduced problems is not possible     

OLS with multiple high dimensional category variables

A new algorithm is proposed for OLS estimation of linear models with multiple high-dimensional category variables. It is a generalization of the within transformation to arbitrary number of category variables. The approach, unlike other fast methods for solving such problems, provides a covariance matrix for the remaining coefficients. The article also sets out a method for solving the resulting sparse system, and the new scheme is shown, by some examples, to be comparable in computational efficiency to other fast methods. The method is also useful for transforming away groups of pure control dummies. A parallelized implementation of the proposed method has been made available as an R-package lfe on CRAN.     

前馈神经网络的混沌学习方法研究

采用混沌优化策略，提出一种前馈神经网络权参数的最优学习方案．由于BP算法优化神经网络权参数时存在收敛速度慢、自身参数选取困难、易陷入局部极小等缺陷．采用混沌变量优化神经网络权参数，具有全局性、快速性、并行性的特点．仿真实验表明采用该方案对强非线性问题的逼近具有精度较高、学习较快的优点．     

Multirate and composite control of two-time-scale discrete-time systems

The design of stabilizing feedback control of singularly perturbed diserete-time systems is decomposed into the design of slow and fast controllers which are combined to form the composite control. Composite control strategies are developed for the case of single rate measurements (all variables are measured at the same rate) as well as for the case of multirate measurements (slow variables are measured at a rate slower than that of fast variables).