Search

This paper considers high-speed control of constrained linear parameter-varying systems using model predictive control. Existing model predictive control schemes for control of constrained linear parameter-varying systems typically require the solution of a semi-definite program at each sampling instance. Recently, variants of explicit model predictive control were proposed for linear parameter-varying systems with polytopic representation, decreasing the online computational effort by orders of magnitude. Depending on the mathematical structure of the underlying system, the constrained finite-time optimal control problem can be solved optimally, or close-to-optimal solutions can be computed. Constraint satisfaction, recursive feasibility and asymptotic stability can be guaranteed a priori by an appropriate selection of the terminal state constraints and terminal cost. The paper at hand gathers previous developments and provides new material such as a proof for the optimality of the solution, or, in the case of close-to-optimal solutions, a procedure to determine a bound on the suboptimality of the solution.

A common assumption when proving stability of linear MPC algorithms fort racking applications is to assume that the desired setpoint is located farinto the interior of the feasible set. The reason for this is that the terminal state constraint set which is centered around the setpoint must be contained within the feasible set. In many applications this assumption can be severly limiting since the terminal set is relatively large and therefore limits how close the setpoint can be to the boundary of the feasible set. We present simple modifications that can be performed in order to guarantee stability and convergence to setpoints located arbitrarily close to the boundary of the feasible set. The main idea is to introduce a scaling variable which dynamically scales the terminal state constraint set and therefore allowsa setpoint to be located arbitrarily close to the boundary. In addition to this the concept of pseudo setpoints are used to gain the maximum possible region of attraction and to handle infeasible references. Recursive feasibility and convergence to the desired setpoint, or its closest feasible alternative, is proven and a motivating example of controlling an agile fighter aircraftis given.

This paper presents the robust optimization framework in the modelling language YALMIP, which carries out robust modelling and uncertainty elimination automatically and allows the user to concentrate on the high-level model. While introducing the software package, a brief summary of robust optimization is given, as well as some comments on modelling and tractability of complex convex uncertain optimization problems.

One of the most fundamental problems in model predictive control (MPC) is the lack of guaranteed stability and feasibility. It is shown how Farkas Lemma in combination with bilevel programming and disjoint bilinear programming can be used to search for problematic initial states which lack recursive feasibility, thus invalidating a particular MPC controller. Alternatively, the method can be used to derive a certificate that the problem is recursively feasible. The results are initially derived for nominal linear MPC, and thereafter extended to the additive disturbance case., Funding Agencies|CENIIT at Linkoping University

Frequent inlet flow changes, especially in the same direction, typically cause problems for averaging level controllers. To obtain optimal flow filtering while being robust towards future inlet flow upsets closed loop robust MPC was used. Its performance and robustness is analyzed and compared to the optimal averaging level controller. The knowledge gained from the robust MPC exercise is also used to propose a robustification of the optimal controller. Both the analysis and the simulation results show that the robust controller obtains comparable flow filtering as the optimal controller even when inlet flow changes are sparse while handling frequent upsets considerably better

In the presence of frequent inlet flow upsets, tuning of averaging level controllers is typically quite complicated since not only the size of the individual steps but also the time in between the subsequent steps need to considered. One structured way to achieve optimal filtering for such a case is to use Robust Model Predictive Control. The robust MPC controller is, however, quite computationally demanding and not easy to implement. In this paper two linear controllers, which mimic the behavior of the robust MPC, are proposed. Tuning guidelines to avoid violation of the tank level constraints as well as to achieve optimal filtering are presented.

Support processes in industrial energy systems, such as heating, ventilation and cooling systems, are important processes in industrial premises as they are related to energy cost, product quality as well as the indoor environment. In the vehicle production process the paint shop is the most energy intensive part, and about 75% of the energy is used in the ovens and spray booths. The spray booth line, which includes paint application and the oven, uses large quantities of air in order to keep the air quality in an optimal range to achieve the desired paint quality. The approach used in paint shops has up to now been to keep as much of steady state conditions as possible to avoid paint defects due to disturbances in the balance. This means that these high air flows are used also at low and non production hours. There is thus a large potential to increase energy efficiency by controlling the air flow and heating without losing the critical balances. This paper will present an initial post-implementation evaluation of the energy efficiency potential and experiences after running this type of system. CFD has been used to investigate the control strategy.

A given explicit piecewise affine representation of an MPC feedback law is approximated by a single polynomial, computed using linear programming. This polynomial state feedback control law guarantees closed-loop stability and constraint satisfaction. The polynomial feedback can be implemented in real time even on very simple devices with severe limitations on memory storage., Funding Agencies|Scientific Grant Agency of the Slovak Republic|1/0079/091/0095/11|Slovak Research and Development Agency|W-0029-07LPP-0092-07

Many control-related problems can be cast as semidefinite programs. Even though there exist polynomial time algorithms and excellent publicly available solvers, the time it takes to solve these problems can be excessive. What many of these problems have in common, in particular in control, is that some of the variables enter as matrix-valued variables. This leads to a low-rank structure in the basis matrices which can be exploited when forming the Newton equations. In this article, we describe how this can be done, and show how our code, called STRUL, can be used in conjunction with the semidefinite programming solver SDPT3. The idea behind the structure exploitation is classical and is implemented in LMI Lab, but we show that when using a modern semidefinite programming framework such as SDPT3, the computational time can be significantly reduced. Finally, we describe how the modelling language YALMIP has been changed in such a way that our code, which can be freely downloaded, can be interfaced using standard YALMIP commands. This greatly simplifies modelling and usage.

The task of generating time optimal trajectories for a six degrees of freedom industrial robot is discussed and an existing convex optimization formulation of the problem is extended to include new types of constraints. The new constraints are speed dependent and can be motivated from physical modeling of the motors and the drive system. It is shown how the speed dependent constraints should be added in order to keep the convexity of the overall problem. A method to, conservatively, approximate the linear speed dependent constraints by a convex constraint is also proposed. A numerical example proves versatility of the extension proposed in this paper., Vinnova's Industry Excellence Center LINK-SIC at Linköoping University

Controller synthesis for linear parameter varying (LPV) systems has received a lot of attention from the control community. This is mainly motivated by the wide range of non-linear dynamical systems that can be approximated by LPV-systems. In this paper a novel method is presented that, by only using local state space models as data, tries to solve the problem of finding a linear parameter varying output-feedback controller. The method uses non-linear programming and a quasi-Newton framework to solve the problem. The great advantages with the proposed method is that it is possible to impose structure in the controller and that you do not need an lpv-model, only state space models for different values of the scheduling parameters. Finally an example is presented to show the potential of the method.

Many control related problems can be cast as semidefinite programs but, even though there exist polynomial time algorithms and good publicly available solvers, the time it takes to solve these problems can be long. Something many of these problems have in common, is that some of the variables enter as matrix valued variables. This leads to a low-rank structure in the basis matrices which can be exploited when forming the Newton equations. In this paper, we describe how this can be done, and show how our code can be used when using SDPT3. The idea behind this is old and is implemented in LMI Lab, but we show that when using a modern algorithm, the computational time can be reduced. Finally, we describe how the modeling language YALMIP is changed in such a way that our code can be interfaced using standard YALMIP commands, which greatly simplifies for the user.

Controller synthesis for linear parameter varying (LPV) systems has received a lot of attention from the control community. This is mainly motivated by the wide range of non-linear dynamical systems that can be approximated by LPV-systems. In this paper a novel method is presented that, by only using local state-space models as data, tries to solve the problem of finding a linear parameter varying output-feedback controller. The method uses non-linear programming and a quasi-Newton framework to solve the problem. The great advantages with the proposed method is that it is possible to impose structure in the controller and that you do not need an LPV-model, only state-space models for different values of the scheduling parameters. Finally an example is presented to show the potential of the method.

In this paper we have formulated the problem offinding an LPV-approximation to a system as an optimization problem. For this optimization problem we have presented two possible ways to solve this. The problem is posed as a model reduction problem and formulated such that it should try to preserve the input-output behavior of the system. In the two examples in the paper the potential of the new methods are shown. We have also shown the benefits of using model reduction techniques to capture the input-output behavior to obtain accurate low order LPV-approximations. One method uses SDP-optimization to solve the problem. SDP optimization has been a hot topic during the last years, but the problem with the SDP method is that it scales badly with the dimension of the problem. Also here it has bilinear constraint swhich makes the problem really difficult. With the other method we try to use a more general nonlinear approach which seem to be more suitable for this problem. For this method we have also calculated a gradient that can be used to apply a descentor Newton-like method to solve the problem.

The task of generating time optimal trajectories for a six degrees of freedom industrial robot is discussed and an existing convex optimization formulation of the problem is extended to include new types of constraints. The new constraints are speed dependent and can be motivated from physical modeling of the motors and the drive system. It is shown how the speed dependent constraints should be added in order to keep the convexity of the overall problem. A method to, conservatively, approximate the linear speed dependent constraints by a convex constraint is also proposed. A numerical example proves versatility of the extension proposed in this paper., Vinnova's Industry Excellence Center LINK-SIC at Linköoping University

Many control related problems can be cast as semidefinite programs but, even though there exist polynomial time algorithms and good publicly available solvers, the time it takes to solve these problems can be long. Something many of these problems have in common, is that some of the variables enter as matrix valued variables. This leads to a low-rank structure in the basis matrices which can be exploited when forming the Newton equations. In this paper, we describe how this can be done, and show how our code can be used when using SDPT3. The idea behind this is old and is implemented in LMI Lab, but we show that when using a modern algorithm, the computational time can be reduced. Finally, we describe how the modeling language YALMIP is changed in such a way that our code can be interfaced using standard YALMIP commands, which greatly simplifies for the user.

This paper addresses the issue of the practical implementation of Model Predictive Controllers (MPC) to processes with short sampling times. Given an explicit solution to an MPC problem, the main idea is to approximate the optimal control law defined over state space regions by a single polynomial of pre-specified degree which, when applied as a state-feedback, guarantees closed-loop stability, constraint satisfaction, and a bounded performance decay. It is shown how to search for such a polynomial by solving a single linear program.

In this paper we have formulated the problem offinding an LPV-approximation to a system as an optimization problem. For this optimization problem we have presented two possible ways to solve this. The problem is posed as a model reduction problem and formulated such that it should try to preserve the input-output behavior of the system. In the two examples in the paper the potential of the new methods are shown. We have also shown the benefits of using model reduction techniques to capture the input-output behavior to obtain accurate low order LPV-approximations. One method uses SDP-optimization to solve the problem. SDP optimization has been a hot topic during the last years, but the problem with the SDP method is that it scales badly with the dimension of the problem. Also here it has bilinear constraint swhich makes the problem really difficult. With the other method we try to use a more general nonlinear approach which seem to be more suitable for this problem. For this method we have also calculated a gradient that can be used to apply a descentor Newton-like method to solve the problem.

Many optimization problems gain from being interpreted and solved in either primal or dual forms. For a user with a particular application, one of these forms is usually much more natural to use, but this is not always the most efficient. This paper presents an implementation in the optimization modelling tool YALMIP that allows the user to define conic optimization problems in a preferred format, and then automatically derive a symbolic YALMIP model of the dual of this problem, solve the dual, and recover original variables. Applications in flexible generation of sum-of-squares programs, and efficient formulations of large-scale experiment design problems are used as illustrative examples.

Checking non-negativity of polynomials using sum-of-squares has recently been popularized and found many applications in control. Although the method is based on convex programming, the optimization problems rapidly grow and result in huge semidefinite programs. Additionally, they often become increasingly ill-conditioned. To alleviate these problems, it is important to exploit properties of the analyzed polynomial, and post-process the obtained solution. This technical note describes how the sum-of-squares module in the MATLAB toolbox YALMIP handles these issues.

We describe a major update of our Matlab freeware GloptiPoly for parsing generalized problems of moments and solving them numerically with semidefinite programming.

For linear and hybrid systems, constrained time-optimal control was shown to be a low complexity alternative to the explicit solution of the constrained finite-time optimal control problem. In this paper we show how Polya's relaxation can be employed to compute minimum-time controllers for discrete-time LPV systems. Contrary to previous publications, our approach allows the use of parameter-varying input matrices. In a comparison over 20 random system, it is shown that compared to explicit LPV-MPC the proposed approach achieves similar or even better control performance, while reducing the complexity of the controller up to an order of magnitude.

In this paper we use bilevel programming to find the maximum difference between a reference controller and a low-complexity controller in terms of the infinity-norm difference of their control laws. A nominal MPC for linear systems with constraints, and a robust MPC for linear systems with bounded additive noise are considered as reference controllers. For possible low-complexity controllers we discuss partial enumeration (PE), Voronoi/closest point, triangulation, linear controller with saturation, and others. A small difference in the norm between a low-complexity controller and a robust MPC may be used to guarantee closed-loop stability of the low-complexity controller and indicate that the behaviour or performance of the low-complexity controller will be similar to that of the reference one. We further discuss how bilevel programming may be used for closed-loop analysis of model reduction.

Recently it was shown how one can reformulate the MPC problem for LPV systems to a series of mpLPs by a closed-loop minimax MPC algorithm based on dynamic programming. A relaxation technique is employed to reformulate constraints which are polynomial in the scheduling parameters to parameter-independent constraints. The algorithm allows the computation of explicit control laws for LPV systems and enables the controller to exploit information about the scheduling parameter. This improves the control performance compared to a standard robust approach where no uncertainty knowledge is used, while keeping the benefits of fast online computations. In this paper the explicit LPV-MPC scheme is extended to incorporate limits on the rate of parameter variation in the controller design. Taking these limits into account can further improve control performance since impossible trajectories of the scheduling parameter do not have to be considered during control law computations.

In Ding et al. [A unified approach for circularity and spatial straightness evaluation using semi-definite programming, International Journal of Machine Tools & Manufacture 47(10) (2007) 1646–1650], the authors advocate semidefinite programming-based relaxations of quadratic optimization problems as a vehicle to solve two circularity and straightness evaluation problems. The purpose of this comment is to point out that the use of semidefinite relaxations for the problems at hand are redundant, since the problems are convex, or changed inter-alia to convex problems. We also take the opportunity to clarify some properties of the semidefinite relaxation, were it to be used for an actual nonconvex problem in this area.

In this paper we propose a closed-loop min-max MPC algorithm based on dynamicprogramming, to compute explicit control laws for systems with a linear parameter-varyingstate transition matrix. This enables the controller to exploit parameter information to improveperformance compared to a standard robust approach where no uncertainty knowledge is used,while keeping the benefits of fast online computations. The off-line computational burden issimilar to what is required for computing explicit control laws for uncertain or nominal LTIsystems. The proposed control strategy is applied to the controlled H´enon map to draw acomparison, in terms of complexity and control performance, with a controller based on apiecewise affine approximation

A considerable amount of optimization problems arising in the control and systemstheory field can be seen as special instances of robust optimization. Much of the modelingeffort in these cases is spent on converting an uncertain problem to a robust counterpartwithout uncertainty. Since many of these conversions follow standard procedures, it is amenableto software support. This paper presents the robust optimization framework in the modelinglanguage YALMIP, which carries out the uncertainty elimination automatically, and allows theuser to concentrate on the high-level model instead., DOI does not work: 10.3182/20080706-5-KR-1001.1729

In this paper we demonstrate how one can reformulate the MPC problem for LPV systems to a series of mpLPs by a closed-loop minimax MPC algorithm based on dynamic programming. A relaxation technique is employed to reformulate constraints which are polynomial in the scheduling parameters to parameter-independent constraints. The algorithm allows the computation of explicit control laws for linear parameter-varying systems and enables the controller to exploit information about the scheduling parameter. This improves the control performance compared to a standard robust approach where no uncertainty knowledge is used, while keeping the benefits of fast online computations. The off-line computational burden is similar to what is required for computing explicit control laws for uncertain or nominal LTI systems. The proposed control strategy is applied to an example to compare the complexity of the resulting explicit control law to the robust controller.

Experiment design involving selection of optimal experiment positions for nonlinear gray-box models is studied. From the derived Fisher information matrix, a convex optimization problem is posed. By considering the dual problem, the experiment design is efficiently solved with linear complexity in the number of candidate positions, compared to cubic complexity for the primal problem. In the numerical illustration, using an industrial robot, the parameter covariance is reduced by a factor of six by using the 15 optimal positions compared to using the optimal single position in all experiments.

This book is part of a three volume set that constitutes the refereed proceedings of the 4th International Symposium on Neural Networks, ISNN 2007, held in Nanjing, China in June 2007. Coverage includes neural networks for control applications, robotics, data mining and feature extraction, chaos and synchronization, support vector machines, fault diagnosis/detection, image/video processing, and applications of neural networks.

Minimax or worst-case approaches have been used frequently recently in model predictive control (MPC) to obtain control laws that are less sensitive to uncertainty. The problem with minimax MPC is that the controller can become overly conservative. An extension to minimax MPC that can resolve this problem is closed-loop minimax MPC. Unfortunately, closed-loop minimax MPC is essentially an intractable problem. In this paper, we introduce a novel approach to approximate the solution to anumber of closed-loop minimax MPC problems. The result is convex optimization problems with size growing polynomially in system dimension and prediction horizon.

A new approach to minimax MPC for systems with bounded external system disturbances and measurement errors is introduced. It is shown that joint deterministic state estimation and minimax MPC can be written as an optimization problem with linear and quadratic matrix inequalities. By linearizing the quadratic matrix inequality, a semidefinite program is obtained. A simulation study indicates that solving the joint problem can improve performance

Robust synthesis is one of the remaining challenges in model predictive control (MPC). One way to robustify an MPC controller is to formulate a minimax problem, i.e., optimize a worst-case performance measure. For systems modeled with an uncertain gain, there are many results available. Typically, the minimax formulations have given intractable problems, or unorthodox performance measures have been used to obtain tractable problems. In this paper, we show how the standard quadratic performance measure can be used in a computationally tractable minimax MPC controller. The controller is developed in a linear matrix inequality framework that easily allows extensions and generalizations.

We present a method to increase feasibility in MPC algorithms that use ellipsoidal terminal state constraints and performance bounds from nominal controllers. The method is based on estimating a bound on the achievable performance with a saturated nominal controller and using this bound in the MPC algorithm. The resulting MPC controller can be implemented efficiently with Second Order Cone Programming.

Almost all proposed approaches to nonlinear receding horizon control with guaranteed stability are based on adding stabilizing constraints, using linearizations of the system, or knowing an upperbound on the value function of an infinite horizon optimal control problem. In this contribution, we present a new approach using none of these ingredients. The idea is to use a dynamic receding horizon controller, where the dynamic part is introduced to tune emphasis between short-range optimality and stability. The result is a design procedure applicable to a nontrivial class of unconstrained nonlinear systems.