Parametric Optimal Control Problems Research Articles

Stabilized weighted reduced order methods for parametrized advection-dominated optimal control problems governed by partial differential equations with random inputs

Abstract In this work, we analyze Parametrized Advection-Dominated distributed Optimal Control Problems with random inputs in a Reduced Order Model (ROM) context. All the simulations are initially based on a finite element method (FEM) discretization; moreover, a space-time approach is considered when dealing with unsteady cases. To overcome numerical instabilities that can occur in the optimality system for high values of the Péclet number, we consider a Streamline Upwind Petrov–Galerkin technique applied in an optimize-then-discretize approach. We combine this method with the ROM framework in order to consider two possibilities of stabilization: Offline-Only stabilization and Offline-Online stabilization. Moreover we consider random parameters and we use a weighted Proper Orthogonal Decomposition algorithm in a partitioned approach to deal with the issue of uncertainty quantification. Several quadrature techniques are used to derive weighted ROMs: tensor rules, isotropic sparse grids, Monte-Carlo and quasi Monte-Carlo methods. We compare all the approaches analyzing relative errors between the FEM and ROM solutions and the computational efficiency based on the speedup-index.

Be greedy and learn: efficient and certified algorithms for parametrized optimal control problems

We consider parametrized linear-quadratic optimal control problems and provide their online-efficient solutions by combining greedy reduced basis methods and machine learning algorithms. To this end, we first extend the greedy control algorithm, which builds a reduced basis for the manifold of optimal final time adjoint states, to the setting where the objective functional consists of a penalty term measuring the deviation from a desired state and a term describing the control energy. Afterwards, we apply machine learning surrogates to accelerate the online evaluation of the reduced model. The error estimates proven for the greedy procedure are further transferred to the machine learning models and thus allow for efficient a posteriori error certification. We discuss the computational costs of all considered methods in detail and show by means of two numerical examples the tremendous potential of the proposed methodology.

AONN-2: An adjoint-oriented neural network method for PDE-constrained shape optimization

PDE-constrained shape optimization has been playing an important role in a large variety of engineering applications. Traditional mesh-dependent shape optimization methods often encounter challenges due to mesh deformation. To address this issue, we present a new adjoint-oriented neural network method, AONN-2, for PDE-constrained shape optimization problems. This method extends the capabilities of the original AONN method [1], which is developed for efficiently solving parametric optimal control problems. AONN-2 inherits from AONN the direct-adjoint looping (DAL) framework for computing the extremum of an objective functional and the involved neural network methods for solving complicated PDEs. Furthermore, AONN-2 expands the application scope to shape optimization by taking advantage of the shape derivatives to optimize the shape represented by discrete boundary points. AONN-2 is a fully mesh-free shape optimization approach, naturally sidestepping issues related to mesh deformation, with no needs for maintaining mesh quality and additional mesh corrections. A series of experimental results are presented, highlighting the flexibility, robustness, and accuracy of AONN-2.

On the Rod Heating Problem by Sources on the Class of Zonal Controls Using the Current and Past State at Measurement Points

We study an optimal feedback control problem for the rod heating process by means of lumped sources. The control actions are the powers of the sources, the values of which are defined on the class of zonal controls. The values of the parameters of zonal control actions are determined by subsets of the state space, to which belong the values of the process state at the measurement points at the current and past time moments. The posed problem is reduced to a parametric optimal control problem on determining a finite-dimensional vector of values of the parameters of zonal control actions. We have obtained optimality conditions for the values of the parameters of zonal control actions. These conditions contain formulas for the gradient of the objective functional with respect to the optimizable parameters. They make it possible to solve the reduced problem numerically with the application of efficient first-order optimization methods.

Https://www.bakumathj.org/archive/Vol2No2/j.bmj.033.pdf

In the paper, we consider an approach for numerical solution to the optimal feedback control problem for an object with distributed parameters on the basis of observation of the object’s phase state at its specific locations by the example of the rod heating process. The control actions are the power of the heat source, the values of which are defined on the class of “zonal” controls. The values of the parameters of zonal control actions are determined by subsets of the state space, to which belong the values of the process state at the measurement points at the current moment of time. The problem of determining zonal controls is reduced to a parametric optimal control problem on determining a finite-dimensional vector of values of the parameters of zonal control actions. We derive optimality conditions for the values of the parameters of zonal control actions. These conditions contain formulas for the gradient of the objective functional with respect to the optimizable parameters of zonal controls. They make it possible to solve the stated problem numerically with the application of efficient first-order optimization methods.

An extended physics informed neural network for preliminary analysis of parametric optimal control problems

An extended physics informed neural network for preliminary analysis of parametric optimal control problems

Sensitivity analysis of the energy balance of dynamic soaring

Dynamic soaring is a periodic flying technique observed in large seabirds that enables the extraction of energy from wind and thereby facilitates covering large distances with minimum effort. It requires wind shears as seen close to the ocean surface or behind mountain ridges. At sufficiently strong winds it becomes feasible to extract at least the amount of energy required for overcoming the aircraft’s dissipation. This situation is often referred to as energy-neutral dynamic soaring and would theoretically enable an abiding effortless flight. An energy-neutral dynamic soaring cycle has already been formulated and solved as a parametric optimal control problem. Due to the inherent variability of the wind, the sensitivity of the energy balance of an open type dynamic soaring cycle to wind conditions will be investigated within this work. Of particular interest is the wind direction and the wind speed. To expose the parameters with the strongest influence for some given conditions, state sensitivities will be computed after the optimization by applying the sensitivity differential equation approach.

Local stability of solutions to a parametric multi-objective optimal control problem

This paper studies local stability of solutions to a parametric multi-objective optimal control problem with constraints. By combining the scalarization method and non-scalarization methods, we show that if the unperturbed problem has a locally strong Pareto solution, then the Pareto solution set is locally Hölder continuous at a reference parameter.

POD-Galerkin model order reduction for parametrized nonlinear time-dependent optimal flow control: an application to shallow water equations

AbstractIn the present paper we propose reduced order methods as a reliable strategy to efficiently solve parametrized optimal control problems governed by shallow waters equations in a solution tracking setting. The physical parametrized model we deal with is nonlinear and time dependent: this leads to very time consuming simulations which can be unbearable, e.g., in a marine environmental monitoring plan application. Our aim is to show how reduced order modelling could help in studying different configurations and phenomena in a fast way. After building the optimality system, we rely on a POD-Galerkin reduction in order to solve the optimal control problem in a low dimensional reduced space. The presented theoretical framework is actually suited to general nonlinear time dependent optimal control problems. The proposed methodology is finally tested with a numerical experiment: the reduced optimal control problem governed by shallow waters equations reproduces the desired velocity and height profiles faster than the standard model, still remaining accurate.

Learning Stochastic Parametric Diferentiable Predictive Control Policies

The problem of synthesizing stochastic explicit model predictive control policies is known to be quickly intractable even for systems of modest complexity when using classical control-theoretic methods. To address this challenge, we present a scalable alternative called stochastic parametric differentiable predictive control (SP-DPC) for unsupervised learning of neural control policies governing stochastic linear systems subject to nonlinear chance constraints. SP-DPC is formulated as a deterministic approximation to the stochastic parametric constrained optimal control problem. This formulation allows us to directly compute the policy gradients via automatic differentiation of the problem's value function, evaluated over sampled parameters and uncertainties. In particular, the computed expectation of the SP-DPC problem's value function is back propagated through the closed-loop system rollouts parametrized by a known nominal system dynamics model and neural control policy which allows for direct model-based policy optimization. We demonstrate the computational efficiency and scalability of the proposed policy optimization algorithm in three numerical examples, including systems with a large number of states or subject to nonlinear constraints.

Solution stability to parametric distributed optimal control problems with finite unilateral constraints

<p style='text-indent:20px;'>This paper deals with stability of solution map to a parametric control problem governed by semilinear elliptic equations with finite unilateral constraints, where the objective functional is not convex. By using the first-order necessary optimality conditions, we derive some sufficient conditions under which the solution map is upper semicontinuous with respect to parameters.</p>

Control variable parameterization and optimization method for stochastic linear quadratic models

The linear quadratic (LQ) optimal control model is widely used in industrial production, medical treatment, finance and other fields. For the stochastic dynamic system, the optimal control of LQ optimal control problem is given in an analytic solution. However, the analytic optimal control is governed by a time-dependent Riccati differential equation, which is often complex and difficult to be solved accurately. Hence, the analytic optimal control may be inconvenient to be implemented in practice. Here, we discuss a parametric optimal control problem of stochastic LQ model. Firstly, the optimal control of a stochastic LQ model is derived. For obtaining an approximate control strategy with a simplified expression, we formulate a parametric stochastic LQ control model, and a control variable parameterization and optimization method is presented to solve optimal control parameter. Finally, for showing effectiveness and practicability of the control variable parameterization and optimization method, an inventory control problem under stochastic environment is discussed.

A weighted POD-reduction approach for parametrized PDE-constrained optimal control problems with random inputs and applications to environmental sciences

Reduced basis approximations of Optimal Control Problems (OCPs) governed by steady partial differential equations (PDEs) with random parametric inputs are analyzed and constructed. Such approximations are based on a Reduced Order Model, which in this work is constructed using the method of weighted Proper Orthogonal Decomposition. This Reduced Order Model then is used to efficiently compute the reduced basis approximation for any outcome of the random parameter. We demonstrate that such OCPs are well-posed by applying the adjoint approach, which also works in the presence of admissibility constraints and in the case of non linear-quadratic OCPs, and thus is more general than the conventional Lagrangian approach. We also show that a step in the construction of these Reduced Order Models, known as the aggregation step, is not fundamental and can in principle be skipped for noncoercive problems, leading to a cheaper online phase. Numerical applications in three scenarios from environmental science are considered, in which the governing PDE is steady and the control is distributed. Various parameter distributions are taken, and several implementations of the weighted Proper Orthogonal Decomposition are compared by choosing different quadrature rules.

Upper semicontinuity of the solution map to a parametric boundary optimal control problem with unbounded constraint sets

The paper studies the solution stability of a parametric control problem governed by semilinear elliptic equations with a mixed state-control constraint, where the objective function is nonconvex and the admissible set is unbounded. We show that under certain conditions, the solution set is upper semicontinuous and continuous with respect to parameters.

OPTIMAL REGULATION OF LOCAL ENERGY SYSTEM WITH RENEWABLE ENERGY SOURCES

The paper is devoted to solving the balancing problem in local power systems with renewable energy sources. For a power system optimization problem, whose operation depends on random weather factors, a convex parameter optimization or optimal control problem was solved using controlled generation, for each individual realization of a random process as a deterministic function, and then statistical processing of results over a set of random realizations was performed and distribution density functions of the desired target function were constructed, followed by estimation of expected values and their confidence intervals. The process describing current deviations of generated power from mean value is modelled as discrete stray model and has properties of Ornstein-Uhlenbeck process, which allowed varying the duration of unit interval, in particular to select data bases of operating objects with inherent temporal discreteness of their monitoring systems. Random components are investigated and modelled, while the average values are considered to be deterministic and are provided within a predictable schedule using also traditional energy sources (centralised power grid). A mathematical model of the combined operation of renewable energy sources in a system with variable load, electric storage device and auxiliary regulating generator is implemented as a scheme of sequential generation and consumption models and random processes describing the current state of the power system. The operation of the electricity accumulators is dependent on the processes mentioned, but in the full balance, it appears together with generation or load losses, which are cumulative sums of unbalanced power and may have a different distribution from the normal one. However, these processes are internal, relating to the redistribution of energy within a generation system whose capacity is generally described satisfactorily, given the relevant criteria, by a normal law. Under this condition, it is possible to estimate the probability of different circumstances - over- or under-generation, that is, to give a numerical estimate of the reliability of energy supply.

Generalized Clarke epiderivatives of the extremum multifunction to a multi-objective parametric discrete optimal control problem

<p style='text-indent:20px;'>This paper deals with the generalized Clarke epiderivative of the extremum multifunction of a multi-objective parametric convex discrete optimal control problem with linear state equations and control constraints. By establishing an abstract result on the generalized epiderivative of the extremum multifunction of a multi-objective parametric convex mathematical programming problem, we derive a formula for computing the generalized Clarke epiderivative of the extremum multifunction to a multi-objective parametric convex discrete optimal control problem. Examples are given to illustrate the obtained results.</p>

Parametric approximate optimal control of uncertain differential game with application to counter terror

Abstract The linear quadratic differential game plays an important role in many fields. It is well known that the saddle point of linear quadratic differential game is given in a feedback form with a solution of Riccati differential equation. However, the control-related Riccati differential equation cannot be solved analytically in many cases. Then the optimal controls may be difficult to be implemented in practice. In order to simplify the forms of optimal controls, in this paper, we investigate a parametric approximate optimal control problem of linear quadratic differential game under uncertain environment. First, we introduce an uncertain linear quadratic differential game model and its analytic optimal controls. Then, an uncertain linear quadratic differential game model with constrained parametric control domain is formulated. Moreover, a parametric approximate optimization method is presented for solving the optimal control parameters. Finally, a counter terror problem is analyzed to show the efficiency of our presented method.

On Hölder calmness and Hölder well-posedness for optimal control problems

In this paper, we consider parametric optimal control problems with linear state equations and control constraints. Applying the KKM fixed point theorem and the Gronwall Lemma, conditions of the existence of solutions to such problems are gained. Under suitable tools and techniques, we formulate sufficient conditions of a Hölder property of solution maps to reference problems without assuming conditions related to differentiability properties. Also, we suggest and investigate a concept of Hölder well-posedness for these problems. Finally, as applications, employing the main results, we discuss the Hölder calmness property of the solution maps to two practical situations, namely linear quadratic regulator and fuel-optimal frictionless horizontal motion of a mass point problems.

Sensitivity Analysis of Multi-objective Optimal Control Problems

Sensitivity analysis of multi-objective parametric optimal control problems with nonconvex cost functions and control constraints is given in this paper. By establishing an abstract result on the Mordukhovich subdifferential of the efficient point multifunction of a multi-objective parametric mathematical programming problem, we derive a formula for computing the Mordukhovich subdifferential of the efficient point multifunction to a multi-objective parametric optimal control problem.

Differential stability of discrete optimal control problems with mixed contraints

Differential stability of parametric discrete optimal control problems in Banach spaces with nonconvex cost functions, nonlinear state equations and mixed constraints is studied. By establishing an abstract result on the subdifferentials of the value function of a parametric mathematical programming problem with geometrical, equality and inequality constraints, we derive formulas for computing the Mordukhovich subdifferential and the singular subdifferential of the value function to a parametric discrete optimal control problem.