## Optimization over state feedback policies for robust control with constraints (2005)

### Cached

### Download Links

Citations: | 33 - 4 self |

### BibTeX

@MISC{Goulart05optimizationover,

author = {Paul J. Goulart and Eric C. Kerrigan and Jan M. Maciejowski},

title = {Optimization over state feedback policies for robust control with constraints},

year = {2005}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper is concerned with the optimal control of linear discrete-time systems, which are subject to unknown but bounded state disturbances and mixed constraints on the state and input. It is shown that the class of admissible affine state feedback control policies with memory of prior states is equivalent to the class of admissible feedback policies that are affine functions of the past disturbance sequence. This result implies that a broad class of constrained finite horizon robust and optimal control problems, where the optimization is over affine state feedback policies, can be solved in a computationally efficient fashion using convex optimization methods without having to introduce any conservatism in the problem formulation. This equivalence result is used to design a robust receding horizon control (RHC) state feedback policy such that the closed-loop system is input-to-state stable (ISS) and the constraints are satisfied for all time and for all allowable disturbance sequences. The cost that is chosen to be minimized in the associated finite horizon optimal control problem is a quadratic function in the disturbance-free state and input sequences. It is shown that the value of the receding horizon control law can be calculated at each sample instant using a single, tractable and convex quadratic program (QP) if the disturbance set is polytopic or given by a 1-norm or ∞-norm bound, or a second-order cone program (SOCP) if the disturbance set is ellipsoidal or given by a 2-norm bound.

### Citations

5102 |
Matrix Analysis
- Horn, Johnson
- 1985
(Show Context)
Citation Context ... ≤ 1} , (27) where E ∈ Rn×l and f ∈ Rn . If this is the case, then an analytical expression for the solution to the row-wise maximization in (25) is easily found. From the definition of the dual norm =-=[24]-=-, it immediately follows that if W is given as in (27), then max w∈W aT w = ‖E T a‖q + a T f, (28) for any vector a ∈ Rn , where 1/p + 1/q = 1. As an example, we give a detailed exposition for the ∞-n... |

4079 | Convex Optimization - Boyd, Vandenberghe - 2004 |

1169 |
Nonlinear systems
- Khalil
- 1996
(Show Context)
Citation Context ... attraction X , and also that all state trajectories are bounded for all bounded disturbance sequences. Furthermore, every trajectory φ(x, k, w(·)) → 0 if w(k) → 0 as k → ∞. The reader is referred to =-=[28, 33, 54]-=- and the references therein for a thorough treatment of ISS. In order to be self-contained, we introduce the following useful result, which is easily proven as in [28, Lem 3.5]: Lemma 2 (ISS-Lyapunov ... |

445 |
Predictive Control with Constraints
- Maciejowski
- 2001
(Show Context)
Citation Context ...al control literature [6–8, 58] and a number of different solutions are available, most of which draw on results from set invariance theory [10], ℓ1 optimal control [16, 18, 52] or predictive control =-=[3, 14, 42, 44, 45]-=-. Predictive control is an optimal control technique that is widely employed in industry, mainly because of its strength in being able to handle constraints in a transparent fashion. In predictive con... |

442 |
Constrained model predictive control: stability and optimality
- Mayne, Rawlings, et al.
- 2000
(Show Context)
Citation Context ...al control literature [6–8, 58] and a number of different solutions are available, most of which draw on results from set invariance theory [10], ℓ1 optimal control [16, 18, 52] or predictive control =-=[3, 14, 42, 44, 45]-=-. Predictive control is an optimal control technique that is widely employed in industry, mainly because of its strength in being able to handle constraints in a transparent fashion. In predictive con... |

253 | Adjustable robust solutions of uncertain linear programs
- Ben-Tal, Goryashko, et al.
(Show Context)
Citation Context ... is possible. In order to show this, we exploit a recent result for solving a specific class of robust optimization problems with hard constraints, called adjustable robust counterpart (ARC) problems =-=[5,23]-=-, where the optimization variables correspond to decisions that can be adjusted as soon as the actual value of the uncertainty becomes available. The authors of [5,23] proposed that instead of solving... |

244 |
Model Predictive Control
- Camacho, Bordons
- 1999
(Show Context)
Citation Context ...al control literature [6–8, 58] and a number of different solutions are available, most of which draw on results from set invariance theory [10], ℓ1 optimal control [16, 18, 52] or predictive control =-=[3, 14, 42, 44, 45]-=-. Predictive control is an optimal control technique that is widely employed in industry, mainly because of its strength in being able to handle constraints in a transparent fashion. In predictive con... |

150 | Linear Controller Design: Limits of Performance. PrenticeHall
- Boyd, Barratt
- 1991
(Show Context)
Citation Context ...e of the tube so as to minimize the conservativeness of the solution. Another possibility for reducing the on-line computational requirements is to compute a single dynamic linear controller off-line =-=[13, 16]-=-, but this approach often results in controllers of very high dimension and can be very conservative in practice. Hence, a popular approach in the predictive control literature is to compute one or mo... |

113 |
Set invariance in control
- Blanchini
- 1999
(Show Context)
Citation Context ...problem that has been studied for some time now in the optimal control literature [6–8, 58] and a number of different solutions are available, most of which draw on results from set invariance theory =-=[10]-=-, ℓ1 optimal control [16, 18, 52] or predictive control [3, 14, 42, 44, 45]. Predictive control is an optimal control technique that is widely employed in industry, mainly because of its strength in b... |

109 | Robust Constrained Model Predictive Control Using
- Kothare, Balakrishnan, et al.
- 1996
(Show Context)
Citation Context ...sequence, the set of feasible decision variables is non-convex, in general. Various proposals have been put forward for modifying this problem so that the set of feasible decision variables is convex =-=[27, 36, 53]-=-, but generally this comes with an increase in the conservativeness of the solution. Hence, an interesting question is whether one can re-parameterize the optimal control problem, where the optimizati... |

98 | Input-to-state stability for discrete-time nonlinear systems
- Jiang, Wang
- 2001
(Show Context)
Citation Context ... attraction X , and also that all state trajectories are bounded for all bounded disturbance sequences. Furthermore, every trajectory φ(x, k, w(·)) → 0 if w(k) → 0 as k → ∞. The reader is referred to =-=[28, 33, 54]-=- and the references therein for a thorough treatment of ISS. In order to be self-contained, we introduce the following useful result, which is easily proven as in [28, Lem 3.5]: Lemma 2 (ISS-Lyapunov ... |

78 |
Constrained Optimal Control Of Linear And Hybrid Systems
- Borrelli
- 2003
(Show Context)
Citation Context ...nstraints. If all the relevant constraint sets are polyhedral and the cost function is suitably chosen, then the solution turns out to be a time-varying piecewise affine state feedback control policy =-=[2, 11, 17, 31, 46, 50]-=-. Unfortunately, the practicality of these results are currently limited to small problem sizes. This is because it is easy to find examples where the complexity of the solution to the optimal control... |

64 |
Robust model predictive control: a survey
- Bemporad, Morari
- 1999
(Show Context)
Citation Context |

57 |
Min-Max Feedback Model Predictive Control for Constrained Linear Systems
- Scokaert, Mayne
- 1998
(Show Context)
Citation Context ...ver arbitrary (nonlinear) feedback policies is particularly difficult if constraints have to be taken into account. Current proposals for achieving this using finite dimensional optimization, such as =-=[51]-=-, are computationally intractable since the size of the optimization problem grows exponentially with the size of the problem data. A number of results are also available that allow for the computatio... |

55 | The ISS philosophy as a unifying framework for stability-like behaviour
- Sontag
- 2000
(Show Context)
Citation Context ... attraction X , and also that all state trajectories are bounded for all bounded disturbance sequences. Furthermore, every trajectory φ(x, k, w(·)) → 0 if w(k) → 0 as k → ∞. The reader is referred to =-=[28, 33, 54]-=- and the references therein for a thorough treatment of ISS. In order to be self-contained, we introduce the following useful result, which is easily proven as in [28, Lem 3.5]: Lemma 2 (ISS-Lyapunov ... |

49 |
Ultimate boundedness control for uncertain discrete-time systems via set-induced Lyapunov functions
- Blanchini
- 1994
(Show Context)
Citation Context ...al, mild technical assumptions, it is easy to compute a K and a polytopic Xf that satisfies Assumption 1 if W is a polytope, an ellipsoid or the affine map of a p-norm ball. The reader is referred to =-=[9,10,35,38,49]-=- and the references therein for details. 6.1 Monotonicity of X sf N and Xdf N We are now in a position to give a sufficient condition under which one can guarantee that X sf df sf N (equivalently, XN ... |

46 |
Min–max control of constrained uncertain discrete-time linear systems
- Bemporad, Borrelli, et al.
- 2003
(Show Context)
Citation Context ...nstraints. If all the relevant constraint sets are polyhedral and the cost function is suitably chosen, then the solution turns out to be a time-varying piecewise affine state feedback control policy =-=[2, 11, 17, 31, 46, 50]-=-. Unfortunately, the practicality of these results are currently limited to small problem sizes. This is because it is easy to find examples where the complexity of the solution to the optimal control... |

46 |
Theory and computation of disturbance invariance sets for discrete-time linear systems
- Kolmanovsky, Gilbert
- 1998
(Show Context)
Citation Context ...w ≤ e for all w ∈ W if and only if sup { aT w | w ∈ W } ≤ e, where a is a vector of appropriate length, e is a scalar and sup { aT w | w ∈ W } is the value of the support function of W evaluated at a =-=[35]-=-; note that the supremum is attained due to W being compact. Hence, one can eliminate the universal quantifier in (23) to obtain the equivalent expression Π df { } (M, v) satisfies (19) N (x) = (M, v)... |

38 |
Minimax approaches to robust model predictive control
- Löfberg
- 2003
(Show Context)
Citation Context ...nally tractable linear program (LP). An equivalent parameterization was also proposed in the literature on predic3tive control with Gaussian state disturbances [55,56] and bounded state disturbances =-=[40,41]-=-. However, the question whether this alternative parameterization is more or less conservative than an affine state feedback parameterization, was not addressed in detail. It was also clear that furth... |

34 | On the minimax reachability of target sets and target tubes - Bertsekas, Rhodes - 1971 |

24 |
Robustifying model predictive control of constrained linear systems
- Mayne, Langson
(Show Context)
Citation Context ...line computation to the selection of one of these control laws (if there are more than one), followed with the computation of a finite sequence of admissible perturbations to the selected control law =-=[1,15,30,38,47]-=-. Though this approach considerably reduces the computational complexity, it is not always obvious how to best choose the linear control laws off-line so as to minimize conservativeness. An obvious im... |

22 |
Constrained receding horizon predictive control for systems with disturbances
- Lee, Kouvaritakis
- 1999
(Show Context)
Citation Context ...line computation to the selection of one of these control laws (if there are more than one), followed with the computation of a finite sequence of admissible perturbations to the selected control law =-=[1,15,30,38,47]-=-. Though this approach considerably reduces the computational complexity, it is not always obvious how to best choose the linear control laws off-line so as to minimize conservativeness. An obvious im... |

21 |
Robust model predictive control using tubes
- Langson, Chryssochoos, et al.
(Show Context)
Citation Context ...control policy. An interesting recent proposal in the predictive control literature is to define a polytopic tube over the control horizon and compute control inputs only at the vertices of this tube =-=[37]-=-. Though this approach is computationally tractable and is equivalent to optimizing over a restricted class of piecewise affine state feedback control policies, it currently suffers from the drawback ... |

20 |
Input-to-state stable MPC for constrained discrete-time nonlinear systems with bounded additive uncertainties
- Limon, Alamo, et al.
- 2002
(Show Context)
Citation Context ...of RHC laws with input constraints only [25,26,34], RHC laws with discontinuities and output feedback [48] and in the analysis and synthesis of RHC laws with robust constraint satisfaction guarantees =-=[30,39,43]-=-. 24Consider a nonlinear, time-invariant, discrete-time system of the form x + = f(x, w), (77) where x ∈ R n is the state and w ∈ R l is a disturbance that takes on values in a compact set W ⊂ R l co... |

19 |
Uncertainty-immunized solutions in linear programming
- Guslitser
- 2002
(Show Context)
Citation Context ... is possible. In order to show this, we exploit a recent result for solving a specific class of robust optimization problems with hard constraints, called adjustable robust counterpart (ARC) problems =-=[5,23]-=-, where the optimization variables correspond to decisions that can be adjusted as soon as the actual value of the uncertainty becomes available. The authors of [5,23] proposed that instead of solving... |

19 | Robust Time-Optimal Control of Constrained - Mayne, Schroeder - 1997 |

18 |
Reducing conservativeness in predictive control of constrained systems with disturbances
- Bemporad
- 1998
(Show Context)
Citation Context ...line computation to the selection of one of these control laws (if there are more than one), followed with the computation of a finite sequence of admissible perturbations to the selected control law =-=[1,15,30,38,47]-=-. Though this approach considerably reduces the computational complexity, it is not always obvious how to best choose the linear control laws off-line so as to minimize conservativeness. An obvious im... |

18 | Robust Dynamic Programming for Min-Max Model Predictive Control of Constrained Uncertain Systems
- Diehl, Björnberg
- 2004
(Show Context)
Citation Context ...nstraints. If all the relevant constraint sets are polyhedral and the cost function is suitably chosen, then the solution turns out to be a time-varying piecewise affine state feedback control policy =-=[2, 11, 17, 31, 46, 50]-=-. Unfortunately, the practicality of these results are currently limited to small problem sizes. This is because it is easy to find examples where the complexity of the solution to the optimal control... |

16 | On robust optimization and the optimal control of constrained linear systems with bounded state disturbances
- Kerrigan, Maciejowski
- 2003
(Show Context)
Citation Context |

16 |
Invariant approximations of the minimal robust positively invariant set
- Rakovic, Kerrigan, et al.
- 2005
(Show Context)
Citation Context ...al, mild technical assumptions, it is easy to compute a K and a polytopic Xf that satisfies Assumption 1 if W is a polytope, an ellipsoid or the affine map of a p-norm ball. The reader is referred to =-=[9,10,35,38,49]-=- and the references therein for details. 6.1 Monotonicity of X sf N and Xdf N We are now in a position to give a sufficient condition under which one can guarantee that X sf df sf N (equivalently, XN ... |

15 | Suciently informative functions and the minimax feedback control of uncertain dynamic systems - Bertsekas, Rhodes - 1973 |

14 |
On decisions rules in stochastic programming
- Gartska, Wets
- 1974
(Show Context)
Citation Context ... each step, i.e. wi = xi+1 − Axi − Bui, ∀i ∈ Z [0,N−1]. (18) The above parameterization appears to have originally been suggested some time ago within the context of stochastic programs with recourse =-=[19]-=-. More recently, it has been revisited as as a means for finding solutions to a class of robust optimization problems, called affinely adjustable robust counterpart (AARC) problems [5,23], and robust ... |

12 | Infinite-time reachability of state-space regions by using feedback control - Bertsekas - 1972 |

12 |
Robust model predictive control of constrained linear systems
- Smith
- 2004
(Show Context)
Citation Context ...sequence, the set of feasible decision variables is non-convex, in general. Various proposals have been put forward for modifying this problem so that the set of feasible decision variables is convex =-=[27, 36, 53]-=-, but generally this comes with an increase in the conservativeness of the solution. Hence, an interesting question is whether one can re-parameterize the optimal control problem, where the optimizati... |

11 |
Approximations of closed-loop MPC
- Löfberg
- 2003
(Show Context)
Citation Context ...nally tractable linear program (LP). An equivalent parameterization was also proposed in the literature on predic3tive control with Gaussian state disturbances [55,56] and bounded state disturbances =-=[40,41]-=-. However, the question whether this alternative parameterization is more or less conservative than an affine state feedback parameterization, was not addressed in detail. It was also clear that furth... |

9 |
Systems with persistent state disturbances: predictive control with restricted constraints
- Chisci, Rossiter, et al.
- 2001
(Show Context)
Citation Context |

9 |
Control of Uncertain Systems
- Dahleh, Diaz-Bobillo
- 1995
(Show Context)
Citation Context ...died for some time now in the optimal control literature [6–8, 58] and a number of different solutions are available, most of which draw on results from set invariance theory [10], ℓ1 optimal control =-=[16, 18, 52]-=- or predictive control [3, 14, 42, 44, 45]. Predictive control is an optimal control technique that is widely employed in industry, mainly because of its strength in being able to handle constraints i... |

8 | Properties of a new parameterization for the control of constrained systems with disturbances
- Kerrigan, Maciejowski
- 2004
(Show Context)
Citation Context ...receding horizon fashion. The paper concludes in Section 8 and suggests some topics for further research. Some of the results in this paper have been published by the authors in the conference papers =-=[21,30,32]-=-. For a detailed discussion on the numerical implementation of the results in this paper, the reader is referred to the conference paper [20] and technical report [22]. Notation and definitions: For m... |

8 |
Discrete-time certainty equivalence output feedback: allowing discontinuous control laws including those from model predictive control
- Messina, Tuna, et al.
- 2005
(Show Context)
Citation Context ...ive is input-to-state stability (ISS), which has already been proven to be effective in the study of RHC laws with input constraints only [25,26,34], RHC laws with discontinuities and output feedback =-=[48]-=- and in the analysis and synthesis of RHC laws with robust constraint satisfaction guarantees [30,39,43]. 24Consider a nonlinear, time-invariant, discrete-time system of the form x + = f(x, w), (77) ... |

7 |
Relationships between affine feedback policies for robust control with constraints
- Goulart, Kerrigan
- 2005
(Show Context)
Citation Context ...receding horizon fashion. The paper concludes in Section 8 and suggests some topics for further research. Some of the results in this paper have been published by the authors in the conference papers =-=[21,30,32]-=-. For a detailed discussion on the numerical implementation of the results in this paper, the reader is referred to the conference paper [20] and technical report [22]. Notation and definitions: For m... |

7 |
A convex parameterization for solving constrained min-max problems with a quadratic cost
- Kerrigan, Alamo
- 2004
(Show Context)
Citation Context ...osing to work with whichever of our two parameterizations is most natural in each context. Before proceeding, we note that alternative cost functions are certainly possible; the reader is referred to =-=[29]-=- where a worst-case quadratic cost function is used and the disturbance is negatively weighted as in H∞ control. 207.1 Cost Function (x) to be one which minimizes the value of a cost function that is... |

5 | Efficient robust optimization for robust control with constraints
- Goulart, Kerrigan, et al.
(Show Context)
Citation Context ...s in the conference papers [21,30,32]. For a detailed discussion on the numerical implementation of the results in this paper, the reader is referred to the conference paper [20] and technical report =-=[22]-=-. Notation and definitions: For matrices A and B, A ⊗ B is the Kronecker product of A and B, A † is the one-sided or pseudo-inverse of A, and A ≤ B denotes element-wise inequality and abs(A) is the el... |

5 | Robust stability of min-max MPC controllers for nonlinear systems with bounded uncertainties
- Limon, Alamo, et al.
- 2004
(Show Context)
Citation Context ...of RHC laws with input constraints only [25,26,34], RHC laws with discontinuities and output feedback [48] and in the analysis and synthesis of RHC laws with robust constraint satisfaction guarantees =-=[30,39,43]-=-. 24Consider a nonlinear, time-invariant, discrete-time system of the form x + = f(x, w), (77) where x ∈ R n is the state and w ∈ R l is a disturbance that takes on values in a compact set W ⊂ R l co... |

4 |
An ecient decompositionbased formulation for robust control with constraints
- Goulart, Kerrigan
- 2005
(Show Context)
Citation Context ...en published by the authors in the conference papers [21,30,32]. For a detailed discussion on the numerical implementation of the results in this paper, the reader is referred to the conference paper =-=[20]-=- and technical report [22]. Notation and definitions: For matrices A and B, A ⊗ B is the Kronecker product of A and B, A † is the one-sided or pseudo-inverse of A, and A ≤ B denotes element-wise inequ... |

4 | On the ISS property for receding horizon control of constrained linear systems
- Jadbabaie, Morse
(Show Context)
Citation Context ...have to use a different notion of stability. One such alternative is input-to-state stability (ISS), which has already been proven to be effective in the study of RHC laws with input constraints only =-=[25,26,34]-=-, RHC laws with discontinuities and output feedback [48] and in the analysis and synthesis of RHC laws with robust constraint satisfaction guarantees [30,39,43]. 24Consider a nonlinear, time-invarian... |

3 |
Inputto-state stabilizing MPC for neturally stable linear systems subject to input constraints
- Kim, Yoon, et al.
- 2004
(Show Context)
Citation Context ...have to use a different notion of stability. One such alternative is input-to-state stability (ISS), which has already been proven to be effective in the study of RHC laws with input constraints only =-=[25,26,34]-=-, RHC laws with discontinuities and output feedback [48] and in the analysis and synthesis of RHC laws with robust constraint satisfaction guarantees [30,39,43]. 24Consider a nonlinear, time-invarian... |

2 |
An LMI approach to robust model predictive control
- Jia, Krogh, et al.
(Show Context)
Citation Context ...sequence, the set of feasible decision variables is non-convex, in general. Various proposals have been put forward for modifying this problem so that the set of feasible decision variables is convex =-=[27, 36, 53]-=-, but generally this comes with an increase in the conservativeness of the solution. Hence, an interesting question is whether one can re-parameterize the optimal control problem, where the optimizati... |

2 |
Optimization of the ℓ∞-induced norm under full state feedback
- Shamma
- 1996
(Show Context)
Citation Context ...died for some time now in the optimal control literature [6–8, 58] and a number of different solutions are available, most of which draw on results from set invariance theory [10], ℓ1 optimal control =-=[16, 18, 52]-=- or predictive control [3, 14, 42, 44, 45]. Predictive control is an optimal control technique that is widely employed in industry, mainly because of its strength in being able to handle constraints i... |

1 | A globally stabilizing receding horizon controller for neutrally stable linear systems with input constraints
- Jadbabaie, Persis, et al.
- 2002
(Show Context)
Citation Context ...have to use a different notion of stability. One such alternative is input-to-state stability (ISS), which has already been proven to be effective in the study of RHC laws with input constraints only =-=[25,26,34]-=-, RHC laws with discontinuities and output feedback [48] and in the analysis and synthesis of RHC laws with robust constraint satisfaction guarantees [30,39,43]. 24Consider a nonlinear, time-invarian... |

1 |
On the piecewise linear nature of constrained min-max model predictive control with bounded uncertainties
- Ramirez, Camacho
- 2003
(Show Context)
Citation Context |