## Low-authority controller design via convex optimization (1999)

### Cached

### Download Links

Venue: | AIAA Journal of Guidance, Control, and Dynamics |

Citations: | 31 - 14 self |

### BibTeX

@ARTICLE{Hassibi99low-authoritycontroller,

author = {Arash Hassibi and Jonathan How and Stephen Boyd},

title = {Low-authority controller design via convex optimization},

journal = {AIAA Journal of Guidance, Control, and Dynamics},

year = {1999},

volume = {22},

pages = {862--872}

}

### Years of Citing Articles

### OpenURL

### Abstract

In this paper we address the problem of low-authority controller (LAC) design. The premise is that the actuators have limited authority, and hence cannot significantly shift the eigenvalues of the system. As a result, the closed-loop eigenvalues can be well approximated analytically using perturbation theory. These analytical approximations may suffice to predict the behavior of the closed-loop system in practical cases, and will provide at least a very strong rationale for the first step in the design iteration loop. We will show that LAC design can be cast as convex optimization problems that can be solved efficiently in practice using interior-point methods. Also, we will show that by optimizing the ℓ1 norm of the feedback gains, we can arrive at sparse designs, i.e., designs in which only a small number of the control gains are nonzero. Thus, in effect, we can also solve actuator/sensor placement or controller architecture design problems. Keywords: Low-authority control, actuator/sensor placement, linear operator perturbation theory, convex optimization, second-order cone programming, semi-definite programming, linear matrix inequality. 1

### Citations

1777 | Atomic decomposition by basis pursuit
- Chen, Donoho, et al.
- 1998
(Show Context)
Citation Context ...nonzero elements of a vector x (subject to some constraints in x) arises in many different fields, but unfortunately, except in very special cases, it is a very difficult problem to solve numerically =-=[26, 27, 28, 29]-=-. However, a relaxation to this problem gives reasonably sparse solutions while being numerically tractable [26, 29]. The method is to minimize the ℓ1 norm of x instead of minimizing its nonzero entri... |

493 | Interior point methods in semidefinite programming with applications to combinatorial optimization
- Alizadeh
- 1995
(Show Context)
Citation Context ...1 + ···+xmFm�G is called a linear matrix inequality (LMI). While SDPs look complicated and would appear difficult to solve, new interior-point methods can solve them with great efficiency (see, e.g., =-=[8, 14]-=-) and several SDP codes are now widely available [15, 16, 17, 18, 19, 20]. The ability to numerically solve SDPs with great efficiency is being applied in several fields, e.g., combinatorial optimizat... |

260 |
Interior Point Polynomial Methods in Convex Programming: Theory and Applications
- Nesterov, Nemirovski
- 1994
(Show Context)
Citation Context ...on we briefly introduce linear, second-order cone, and semi-definite programs which are families of convex optimization problems that can be efficiently solved (globally) using interior-point methods =-=[7, 8]-=-. In later sections, we will see how LAC design can be cast in terms of linear, second-order cone, or semi-definite programs and hence solved efficiently in practice. A linear program (LP) is an optim... |

155 | Linear programming: foundations and extensions
- Vanderbei
- 1996
(Show Context)
Citation Context ...ty for Bc, Cc, andDcwe can find designs that require a small number of actuators and sensors. Another problem that can be formulated in the LAC framework is that of structural design and optimization =-=[6]-=-. In such a case, x can include various parameters such as beam widths, beam lengths, masses, dampers, etc. The best design, for example, is a structure that supports specified loads at fixed points, ... |

151 | a C library for semidefinite programming
- Borchers, “CSDP
- 1999
(Show Context)
Citation Context ...). While SDPs look complicated and would appear difficult to solve, new interior-point methods can solve them with great efficiency (see, e.g., [8, 14]) and several SDP codes are now widely available =-=[15, 16, 17, 18, 19, 20]-=-. The ability to numerically solve SDPs with great efficiency is being applied in several fields, e.g., combinatorial optimization and control [21]. SDP is currently a highly active research area. A s... |

133 |
Applications of second-order cone programming. Linear Algebra and its Applications
- Lobo, Vandenberghe, et al.
- 1998
(Show Context)
Citation Context ...econd-order cone programming (SOCP) for eigenvalue-placement specifications. LPs and SOCPs are convex optimization problems and can be solved very efficiently, both in theory and practice (see, e.g., =-=[4, 5, 6, 7]-=-). 3.1 First order perturbation formulas for eigenvalues of a matrix Consider the family of operators A(x)sR nXn where A(0) = A and xsR q is a parameter supposed to be small. A question arises whether... |

98 | M.: SDPA (SemiDefinite Programming Algorithm) – user’s manual – version 5.00
- Fujisawa, Nakata, et al.
- 1999
(Show Context)
Citation Context ...). While SDPs look complicated and would appear difficult to solve, new interior-point methods can solve them with great efficiency (see, e.g., [8, 14]) and several SDP codes are now widely available =-=[15, 16, 17, 18, 19, 20]-=-. The ability to numerically solve SDPs with great efficiency is being applied in several fields, e.g., combinatorial optimization and control [21]. SDP is currently a highly active research area. A s... |

91 | Primal-dual potential reduction method for problems involving matrix inequalities
- VANDENBERGHE, BOYD
(Show Context)
Citation Context ...on we briefly introduce linear, second-order cone, and semi-definite programs which are families of convex optimization problems that can be efficiently solved (globally) using interior-point methods =-=[7, 8]-=-. In later sections, we will see how LAC design can be cast in terms of linear, second-order cone, or semi-definite programs and hence solved efficiently in practice. A linear program (LP) is an optim... |

79 |
A Short Introduction to Perturbation Theory for Linear Operators
- Kato
- 1982
(Show Context)
Citation Context ...eter supposed to be small. A question arises whether the eigenvalues of A(x) can be expressed as a power series in x, i.e., whether they are holomorphic functions of x in the neighborhood of x =0. In =-=[25]-=- it is shown that if A(x) isk-times continuously differentiable in x on a simplyconnected domain D⊂R q , and the number of eigenvalues λi(x) ofA(x) corresponding to a Jordan block of size 1 is constan... |

53 |
D1993E: “Industrial application of model based predictive control
- Richalet
(Show Context)
Citation Context ..., for example, Zadeh and Whalen observed in 1962 that certain minimum-time and minimum-fuel optimal control problems could be (numerically) solved by linear programming [9]. In the late 70s, Richalet =-=[10]-=- developed model predictive control (also known as dynamic matrix control or receding horizon control), in which linear or quadratic programs are used to solve an optimal control problem at each time ... |

46 |
On the rank minimization problem over a positive semidefinite linear matrix inequality
- Mesbahi, Papavassilopoulos
- 1997
(Show Context)
Citation Context ...nonzero elements of a vector x (subject to some constraints in x) arises in many different fields, but unfortunately, except in very special cases, it is a very difficult problem to solve numerically =-=[26, 27, 28, 29]-=-. However, a relaxation to this problem gives reasonably sparse solutions while being numerically tractable [26, 29]. The method is to minimize the ℓ1 norm of x instead of minimizing its nonzero entri... |

41 | sdpsol: A parser/solver for semidefinite programming and determinant maximization problems with matrix structure. User’s guide
- Wu, Boyd
- 1996
(Show Context)
Citation Context ...). While SDPs look complicated and would appear difficult to solve, new interior-point methods can solve them with great efficiency (see, e.g., [8, 14]) and several SDP codes are now widely available =-=[15, 16, 17, 18, 19, 20]-=-. The ability to numerically solve SDPs with great efficiency is being applied in several fields, e.g., combinatorial optimization and control [21]. SDP is currently a highly active research area. A s... |

36 |
SP: Software for Semidefinite Programming. User’s Guide
- Vandenberghe, Boyd
- 1994
(Show Context)
Citation Context |

31 |
Control system synthesis via bilinear matrix inequalities
- Safonov, Goh, et al.
- 1994
(Show Context)
Citation Context ...on, multiobjective controller synthesis, etc., can be cast as matrix inequalities that are bilinear in the variables (bilinear matrix inequalities or BMIs). See, for example, Safonov, Goh, and others =-=[31, 32, 33, 34, 35]-=-. BMIs are hard to solve directly so people have been looking at a variety of iterative schemes to solve them. The idea of linearizing matrix inequalities of §8 can be used as a path-following (homoto... |

30 | PCx user guide
- Czyzyk, Mehrotra, et al.
- 1996
(Show Context)
Citation Context ...econd-order cone programming (SOCP) for eigenvalue-placement specifications. LPs and SOCPs are convex optimization problems and can be solved very efficiently, both in theory and practice (see, e.g., =-=[4, 5, 6, 7]-=-). 3.1 First order perturbation formulas for eigenvalues of a matrix Consider the family of operators A(x)sR nXn where A(0) = A and xsR q is a parameter supposed to be small. A question arises whether... |

25 |
Safonov and G.P Papavassilopoulos, A global optimization approach for the BMI problem
- Goh, G
- 1994
(Show Context)
Citation Context ...on, multiobjective controller synthesis, etc., can be cast as matrix inequalities that are bilinear in the variables (bilinear matrix inequalities or BMIs). See, for example, Safonov, Goh, and others =-=[31, 32, 33, 34, 35]-=-. BMIs are hard to solve directly so people have been looking at a variety of iterative schemes to solve them. The idea of linearizing matrix inequalities of §8 can be used as a path-following (homoto... |

23 |
LOQO user’s manual
- Vanderbei
- 2000
(Show Context)
Citation Context ...lem at each time step. Model predictive control is now widely used in the process control industry. Several high quality, efficient implementations of interior-point LP solvers areavailable(see,e.g., =-=[11, 12, 13]-=-). A semi-definite program (SDP), is an optimization problem which has the form minimize cT x subject to x1F1 + ···+xmFm�G, Ax = b where Fi and G are symmetric p×p matrices, and the inequality � denot... |

21 |
A new algorithm for computing sparse solutions to linear inverse problems
- Harikumar, Btesler
- 1996
(Show Context)
Citation Context ...nonzero elements of a vector x (subject to some constraints in x) arises in many different fields, but unfortunately, except in very special cases, it is a very difficult problem to solve numerically =-=[26, 27, 28, 29]-=-. However, a relaxation to this problem gives reasonably sparse solutions while being numerically tractable [26, 29]. The method is to minimize the ℓ1 norm of x instead of minimizing its nonzero entri... |

15 |
Vibration Damping of Structural Elements
- Sun, Lu
- 1995
(Show Context)
Citation Context ...e, could be a linear-quadratic-Gaussian (LQG) controller using a collection of sensors and actuators. LAC, however, is usually implemented using (active or passive) high-energy-dissipating mechanisms =-=[5]-=-. High-energy-dissipating mechanisms are usually incorporated into the structure by using layers of viscoelastic shear damping material. In the simplest case, the force-extension characteristic of vis... |

14 |
sdpsol: A Parser/Solver for Semide nite Programming and Determinant Maximization Problems with Matrix Structure. User's Guide, Version Beta
- Wu, Boyd
- 1996
(Show Context)
Citation Context ...s are very powerful and enable us to formulate design objectives beyond eigenvalue-placement speci cations in terms of semide nite programs (SDPs), which can then be solved very e ciently (see, e.g., =-=[9, 10, 11]-=-). These design objectives can be combined to get, for example, a desired eigenvalue location for the system while providing a bound on output energy, L2 gain, etc. Here, the method is illustrated on ... |

13 |
Biaffine matrix inequality properties and computational methods
- GOH, TURAN, et al.
- 1994
(Show Context)
Citation Context ...on, multiobjective controller synthesis, etc., can be cast as matrix inequalities that are bilinear in the variables (bilinear matrix inequalities or BMIs). See, for example, Safonov, Goh, and others =-=[31, 32, 33, 34, 35]-=-. BMIs are hard to solve directly so people have been looking at a variety of iterative schemes to solve them. The idea of linearizing matrix inequalities of §8 can be used as a path-following (homoto... |

12 |
LMI Lab: A Package for Manipulating and Solving LMIs. INRIA
- Gahinet, Nemirovskii
- 1993
(Show Context)
Citation Context |

11 |
Controller synthesis via bilinear matrix inequali- ties
- Safanov, Goh, et al.
- 1994
(Show Context)
Citation Context |

10 | socp: Software for SecondOrder Cone Programming
- Lobo, Vandenberghe, et al.
- 1997
(Show Context)
Citation Context ...econd-order cone programming (SOCP) for eigenvalue-placement specifications. LPs and SOCPs are convex optimization problems and can be solved very efficiently, both in theory and practice (see, e.g., =-=[4, 5, 6, 7]-=-). 3.1 First order perturbation formulas for eigenvalues of a matrix Consider the family of operators A(x)sR nXn where A(0) = A and xsR q is a parameter supposed to be small. A question arises whether... |

10 | A new primal-dual interior-point method for semide programming - Alizadeh, Haeberly, et al. - 1994 |

7 |
Theory of control of structures by lowauthority controllers
- Aubrun
- 1980
(Show Context)
Citation Context ...y Stanford, CA 94305-9510, USA Introduction The premise in low-authority control (LAC) is that the actuators have limited authority, and hence cannot significantly shift the eigenvalues of the system =-=[1, 2]-=-. As a result, the closed-loop eigenvalues can be well approximated analytically using perturbation theory. These analytical approximations may suffice to predict the behavior of the closed-loop syste... |

7 |
On optimal control and linear programming
- Zadeh
- 1962
(Show Context)
Citation Context ...e variety of fields. In control, for example, Zadeh and Whalen observed in 1962 that certain minimum-time and minimum-fuel optimal control problems could be (numerically) solved by linear programming =-=[9]-=-. In the late 70s, Richalet [10] developed model predictive control (also known as dynamic matrix control or receding horizon control), in which linear or quadratic programs are used to solve an optim... |

7 |
sdppack User’s Guide, Version 0.9 Beta. NYU
- Alizadeh, Haeberly, et al.
- 1997
(Show Context)
Citation Context ... programming as special cases, but can also be used to solve a variety of nonlinear, nondifferentiable problems; see, e.g., [22]. Moreover, efficient interior-point software for SOCP is now available =-=[23, 24]-=-. As a final note, it should be mentioned that among the three different class of optimization problems mentioned, SDP is the most general, and includes LP and SOCP as special cases. 5 (4) (5)s5 Eigen... |

7 |
Matrix cones, complementarity problems, and the bilinear matrix inequality,” Decision and Control
- Mesbahi, Papavassilopoulos, et al.
- 1995
(Show Context)
Citation Context |

4 |
A Parser/Solver for Semidefinite Programming and Determinant Maximization Problems with Matrix Structure. User’s Guide, Beta Version
- Wu, Boyd
- 1996
(Show Context)
Citation Context ...e very powerful and enable us to formulate design objectives beyond eigenvalue-placement specifications in terms of semidefinite programs (SDPs), which can then be solved very efficiently (see, e.g., =-=[9, 10, 11]-=-). These design objectives can be combined to get, for example, a desired eigenvalue location for the system while providing a bound on output energy, E2 gain, etc. Here, the method is illustrated on ... |

3 |
Large space structures control: An integrated approach
- Aubrun, Gupta, et al.
- 1979
(Show Context)
Citation Context ...r performance measures beyond eigenvalue-placement. 2 Applications of LAC A key control design methodology for flexible systems with many elastic modes follows the two-level architecture presented in =-=[1, 3]-=-. This architecture consists of a wide-band, low-authority control (LAC) and a narrowband, high-authority control (HAC). Within this framework, the HAC is designed based on a (low-order) finitedimensi... |

3 |
Transmission Zeros and High-Authority/Low-Authority Control of Flexible Space Structures
- Williams
- 1994
(Show Context)
Citation Context ...performance measures beyond eigenvalue-placement. 2s3 Applications of LAC A key control design methodology for flexible systems with many elastic modes follows the two-level architecture presented in =-=[1, 3, 2, 4]-=-. This architecture consists of a wide-band, low-authority control (LAC) and a narrow-band, high-authority control (HAC). Within this framework, the HAC is designed based on a (low-order) finite-dimen... |

3 |
Affine scaling transformation based methods for computing low complexity sparse solutions
- Rao, Gorodnitsky
- 1996
(Show Context)
Citation Context |

2 |
Structural dynamic stability augmentation and gust alleviation of flexible aircraft
- Wykes
- 1968
(Show Context)
Citation Context ...y Stanford, CA 94305-9510, USA Introduction The premise in low-authority control (LAC) is that the actuators have limited authority, and hence cannot significantly shift the eigenvalues of the system =-=[1, 2]-=-. As a result, the closed-loop eigenvalues can be well approximated analytically using perturbation theory. These analytical approximations may suffice to predict the behavior of the closed-loop syste... |

2 |
User's guide to LIPSOL: a matlab toolkit for linear programming interior-point solvers
- Zhang
- 1994
(Show Context)
Citation Context |

2 |
Robust Placement of Actuators and Dampers for Structural Control
- Anderson, Hagood
- 1993
(Show Context)
Citation Context ... Robust actuator/sensor placement and robust controller design are usually performed in two separate stages (and hence non-optimally) because it is numerically intractable to do otherwise (see, e.g., =-=[30]-=- for a thorough overview of robust actuator and damper placement for structural control). It is yet another numerical advantage of LAC design that it is possible to handle both of these problems at on... |

2 | Robust Control Synthesis via Bilinear Matrix Inequalities - Gob - 1995 |

1 |
A Short Introduction to Perturbation Theory for Linear Operators
- Karo
- 1982
(Show Context)
Citation Context ...ter supposed to be small. A question arises whether the eigenvalues of A(x) can be expressed as a power series in x, i.e., whether they are holomorphic functions of x in the neighborhood ofx -- 0. In =-=[8]-=- it is shown that if A(x) is continuously differentiable in x on a simply-connected domain D C R q, and the number of eigenvalues Ai(x) of A(x) corresponding to a Jordan block of size I is constant fo... |

1 |
LMI Lab: A Package for Manipulating and Solving LMIs. INPdA
- Gahinet, Nemirovskii
- 1993
(Show Context)
Citation Context ...e very powerful and enable us to formulate design objectives beyond eigenvalue-placement specifications in terms of semidefinite programs (SDPs), which can then be solved very efficiently (see, e.g., =-=[9, 10, 11]-=-). These design objectives can be combined to get, for example, a desired eigenvalue location for the system while providing a bound on output energy, E2 gain, etc. Here, the method is illustrated on ... |

1 | SDPSOL: A ParserJSolver for Semidefinite Programming and Determinant Maximization Problems with Matrix Structure. User's Guide, Version Beta - Wu, Boyd - 1996 |

1 | LMI Lab: A Package for Manipulating and Solving LMIs. INRIA - Gahinct, Ncmirovskii - 1993 |

1 | Atomic dccomposkion of basis pursuit - Chcn, Donoho, et al. - 1996 |

1 | On the rank minimization problem over a poskivc scmidcfinkc linear matrix inequality - Mcsbahi, Papavassilopoulos - 1997 |

1 | Biafiqnc matrL: inequality properties and computational methods - Gob, Turan, et al. - 1994 |

1 | Matrix cones, complcmcntarky problems and the bilinear matrix inequality - Mcsbahi, Papavassilopoulos, et al. - 1995 |