Abstract
The history of constrained optimization spans nearly three centuries. The principal warhorse, Lagrange multipliers, was discovered by Lagrange in the Statics section of his famous book on Mechanics from 1788, by applying the idea of virtual velocities to problems in statics with constraints. The idea of virtual velocities, in turn, goes back to a letter of Johann Bernoulli from 1715 to Varignon, in which he announced a very simple rule for solving hundreds of Varignon’s problems in the blink of an eye. Varignon then explains this rule in his book published in 1725. Half a century later, Bernoulli’s rule was chosen by Lagrange as the general principle for the foundation of his mechanics, with the multipliers as the main tool for treating mechanical constraints. In the second edition of his mechanics, published in 1811, Lagrange stressed the importance of his multipliers also for constrained optimization. In particular, they provide spectacular simplifications of entire chapters of Euler’s treatise on Variational Calculus from 1744. Lagrange multipliers is however a much farther reaching concept; we show how one can discover the important primal and dual equations in optimal control and the famous maximum principle of Pontryagin using only Lagrange multipliers. Pontryagin and his group, however, did not discover the maximum principle this way, since they were coming from a completely different area of mathematics. We finally give the complete formulation of PDE constrained optimization based on duality introduced by Lions, and conclude with an outlook on more recent applications.
Mathematics Subject Classification (2010). Primary 01-02; Secondary 49-03, 65K10
Our intention is not to write a full historical paper, but to highlight the parts of the historical development we find interesting as mathematicians. For full details on the history of constrained optimization with complete references, see [45] and [46].
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
On the frontispiece is written “Dont le projet fut donné en M.DC.LXXXVII”.
- 2.
“Votre projet d’une nouvelle mechanique fourmille d’un grand nombre d’exemples, dont quelques uns à en juger par les figures paroissent assez compliqués; mais je vous deffie de m’en proposer un à votre choix, que je ne resolve sur le champ et comme en jouant par ma dite regle.”
- 3.
“…cependant permettez moy que je vous reproche ici une nonchalance qui vous est arrivé assez souvent en ce que vous portez quelques fois votre jugement un peu à la legere, sans examiner, si ce que vous croyez etre une objection en est veritablement une; …c’est donc pour une autre fois que je vous donne cet avertissement à fin que vous soyez à l’avenir sur vos gardes, quand il s’agit de juger…”
- 4.
Varignon gave in his book the wrong date 1717, which was also copied by Lagrange.
- 5.
dp, dq, dr are not independent at the equilibrium point.
- 6.
Up to now, we have preserved all letters exactly as they appear in Lagrange, but we have changed this potential, denoted \(\Pi \) by Lagrange, to U, as it is usual now.
- 7.
See also Carathéodory [16] for a general study of equivalent formulations.
- 8.
- 9.
- 10.
Personal communication of Plail with Boltyanski, and explanation by Gamkrelidze in his paper about the discovery of the maximum principle:
- 11.
In fact, since the endpoint is fixed as well, no variations are allowed at the endpoint either, but then Pontryagin could not have obtained the solution (3.43) of the then overdetermined system of ordinary differential equations (3.42), and thus he decided to first only fix the starting point [27, p. 442]. This flaw was only later fixed by Boltyanski, see the end of this subsection.
- 12.
To solve the time optimal control problem correctly using Lagrange multipliers, we need to introduce the time variable as a state variable, y 0(t): = t, which implies \(\dot{y}_{0} = 1\), y 0(0) = 0. The correct Lagrangian then becomes \(\mathcal{L}(\boldsymbol{y},\boldsymbol{\lambda },\boldsymbol{u}) = y_{0}(T) +\int _{ 0}^{T}\boldsymbol{\lambda }^{T}(\dot{\boldsymbol{y}} -\boldsymbol{ g}(\boldsymbol{y},\boldsymbol{u}))\mathit{dt}\), where all vectors are now one element longer. Computing the variational derivative with respect to \(\boldsymbol{y}\), we obtain now in addition to the earlier equations \(\dot{\lambda }_{0} = 0\) and \(z_{0}(T) +\lambda _{0}(T)z_{0}(T) = 0\) for arbitrary variation z 0, which implies \(\lambda _{0}(T) = -1\) and hence \(\lambda _{0}(t) = -1\) to complete the time optimality system with y 0(t): = t.
- 13.
See also Footnote 12.
- 14.
On the following pages we will solve the general problem of variational calculus in an (n+1) dimensional space with p ordinary differential equations as constraints, using the method of geodesic equal distances.
- 15.
Das Hauptresultat besteht darin, dass unsere Gefällkurven mit den Cauchyschen Charakteristiken zusammenfallen und Lösungen der kanonischen Differentialgleichungen (3.62) sind, die in der Mechanik eine so bedeutende Rolle spielen.
- 16.
According to J.-L. Lions: “Le travail de Yu. V. Egorov contient une étude détaillée de ce problème, mais nous n’avons pas pu comprendre tous les points des démonstrations de cet auteur, les résultats étant très probablement tous corrects.”
- 17.
Institut de Recherche en Informatique et Automatique, the precursor of the modern INRIA.
- 18.
“La formulation (1.31) peut être considérée comme un analogue du principe du maximum de Pontryagin, pour lequel nous référons […] à Pontryagin-Boltyanski-Gamkrelidze-Mischenko” [37].
- 19.
According to J. Blum, it was R. Glowinski, one of the former students of Lions, who once showed Lions on the board that the adjoint state can simply be interpreted as a Lagrange multiplier. This was confirmed by R. Glowinski (personal communication).
References
Archimedes, On the Equilibrium of Planes. publ. Basel (Latin-Greek, 1544), Paris (Latin-Greek, 1615, p.145), Heath (Engl., 1897, p.189), Ver Eecke (French, 1921, I, p.237–299), 250 B.C., in Opera of Archimedes
R. Bellman, I. Glicksberg, O. Gross, On the ’bang-bang’ control problem. Technical Report, DTIC Document, 1955
M. Benzi, G.H. Golub, J. Liesen, Numerical solution of saddle point problems. Acta Numerica, 14(1), pp. 1–137 (2005)
J. Bernoulli, Opera Omnia, 4 vols. (Bousquet & Socios., Lausannae/Genevae, 1742)
G.A. Bliss, Lectures on the Calculus of Variations, vol. 850 (University of Chicago Press, Chicago, 1946)
H.G. Bock, Randwertproblemmethoden zur Parameteridentifizierung in Systemen nichtlinearer Differentialgleichungen. Ph.D. thesis, Rheinische Friedrich-Wilhelms-Universität, Bonn, 1987 (No. 183)
V.G. Boltyanski, The maximum principle in the theory of optimal processes (Russian). Doklady AN SSSR 119(6), 1070–1073 (1958)
V.G. Boltyanski, The Maximum Principle—How it Came to Be? (Inst. für Mathematik, Technische Univ. München, München, 1994)
V.G. Boltyanski, R.V. Gamkrelidze, L.S. Pontryagin, On the theory of optimal processes (Russian). Doklady AN SSSR 110, 7–10 (1956)
V.G. Boltyanski, R.V. Gamkrelidze, L.S. Pontryagin, The theory of optimal processes. I. The maximum principle. Izv. Akad. Nauk SSSR. Ser. Mat. 24, 3–42 (1960)
V. Boltyanski, H. Martini, V. Soltan, Geometric Methods and Optimization Problems, vol. 4 (Springer, New York, 1999)
A. Borzì, V. Schulz, Computational Optimization of Systems Governed by Partial Differential Equations. Computational Science & Engineering (SIAM, Philadelphia, 2012)
D.W. Bushaw, Differential equations with a discontinuous forcing term. Ph.D. thesis, Department of Mathematics, Princeton University, 1952
D.W. Bushaw, Experimental towing tank. Technical Report, Stevens Institute of Technology, Reprint 169, Hoboken, 1953
C. Carathéodory, Über die diskontinuierlichen Lösungen in der Variationsrechnung. Ph.D. thesis, Universität Göttingen, 1904 (Gesammelte Mathematische Schriften, Band I), pp. 1–71
C. Carathéodory, Die Methode der geodätischen Äquidistanten und das Problem von Lagrange. Acta Mathematica 47(3), 199–236 (1926)
T. Carraro, M. Geiger, R. Rannacher, Indirect multiple shooting for nonlinear parabolic optimal control problems with control constraints. SIAM J. Sci. Comput. 36(2), A452–A481 (2014)
Y.V. Egorov, Some problems in the theory of optimal control. Dokl. Akad. Nauk. SSSR 145, 720–723 (1962)
Y.V. Egorov, Sufficient conditions for optimal control in Banach spaces. Mat. Sbornik 64, 79–101 (1964)
L. Euler, Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes sive solutio problematis isoperimetrici latissimo sensu accepti (Bousquet & Socios., Lausannae/Genevae, 1744) [Enestr. 65, Opera Omnia, Ser.I, vol. 24]
H.O. Fattorini, Time-optimal control of solutions of operational differential equations. J. SIAM Control Ser. A 2(1), 54–59 (1964)
A.A. Feldbaum, The simplest relay system of automatic control (Russian). Avtomatika i Telemehanika 10(4), 249–266 (1949)
A.A. Feldbaum, Optimal processes in systems of automatic control (Russian). Avtomatika i Telemehanika 14(6), 712–728 (1953)
A.A. Feldbaum, On synthesis of optimal systems of automatic control (Russian), in Transactions of the 2nd National Conference on the Theory of Automatic Control, Izdat. AN SSSR, vol. 2 (1955), pp. 325–360
A.A. Feldbaum, On synthesis of optimal systems with the aid of phase space (Russian). Avtomatika i Telemehanika 16(2), 129–149 (1955)
A. Friedman, Optimal control for parabolic equations. J. Math. Anal. Appl. 18, 479–491 (1967)
R.V. Gamkrelidze, Discovery of the maximum principle. J. Dyn. Control Syst. 5(4), 437–451 (1999)
M.J. Gander, G. Wanner, From Euler, Ritz and Galerkin to modern computing. SIAM Rev. 54, 627–666 (2013)
M. Heinkenschloss, A time-domain decomposition iterative method for the solution of distributed linear quadratic optimal control problems. J. Comput. Appl. Math. 173, 169–198 (2005)
H.K. Hesse, G. Kanschat, Mesh adaptive multiple shooting for partial differential equations. Part I: linear quadratic optimal control problems. J. Numer. Math. 17(3), 195–217 (2009)
M.R. Hestenes, A general problem in the calculus of variations with applications to paths of least time. Technical Report, RAND Memorandum RM-100, 1950. ASTIA Document Number AD 112382
H.B. Keller, Numerical Methods for Two-Point Boundary Value Problems (Waltham, Blaisdell, 1968)
J.L. Lagrange, Méchanique analitique (Chez la Veuve Desaint, A Paris, 1788)
J.L. Lagrange, Mécanique analytique. (Mme Ve Courcier, Paris, 1811/1815) [Second enlarged edition in two volumes; third edition 1853 publ. by J. Bertrand; fourth edition in Oeuvres de Lagrange, vol. 11,12, 1888]
A. Lerner, Improving of dynamic properties of automatic compensators with the aid of nonlinear connections I (Russian). Avtomatika i Telemehanika 13(2), 134–144 (1952)
A. Lerner, Constructing of time-optimal systems of automatic control with constrained values of coordinates of controled object (Russian), in Transactions of the 2nd National Conference on the Theory of Automatic Control, Izdat. AN SSSR, vol. 2 (1955), pp. 305–324
J.L. Lions, Contrôle optimal de systèmes gouvernés par des équations aux dérivées partielles (Dunod, Paris, 1968)
J.L. Lions, Exact controllability, stabilization and perturbations for distributed systems. SIAM Rev. 30, 1–68 (1988)
S. Mac Lane, P.L. Duren, R.A. Askey, U.C. Merzbach, Mathematics at the University of Chicago: A Brief History (American Mathematical Society, Providence, 1989)
E.J. McShane, On multipliers for Lagrange problems. Am. J. Math. 61(4), 809–819 (1939)
D.D. Morrison, J.D. Riley, J.F. Zancanaro, Multiple shooting method for two-point boundary value problems. Commun. ACM 5(12), 613–614 (1962)
M.R. Osborne, On shooting methods for boundary value problems. J. Math. Anal. Appl. 27(2), 417–433 (1969)
H.J. Pesch, Carathéodory’s royal road of the calculus of variations: Missed exits to the maximum principle of optimal control theory. Numer. Algebra Control Optim. 3(1), 161–173 (2013)
H.J. Pesch, R. Bulirsch, The maximum principle, Bellman’s equation, and Carathéodory’s work. J. Optim. Theory Appl. 80(2), 199–225 (1994)
H.J. Pesch, M. Plail, The maximum principle of optimal control: a history of ingenious ideas and missed opportunities. Control Cybern. 38(4A), 973–995 (2009)
M. Plail, Die Entwicklung der optimalen Steuerungen: von den Anfängen bis zur eigenständigen Disziplin in der Mathematik (Vandenhoeck und Ruprecht, Göttingen, 1998)
L.S. Pontryagin, Optimal regulation processes. Uspekhi Matematicheskikh Nauk 14(1), 3–20 (1959)
L.S. Pontryagin, V.G. Boltyanski, R.V. Gamkrelidze, E.F. Mishchenko, The Mathematical Theory of Optimal Processes (Interscience Publishers/Wiley, New York, 1962)
R. Serban, S. Li, L.R. Petzold, Adaptive algorithms for optimal control of time-dependent partial differential-algebraic equation systems. Int. J. Numer. Methods Eng. 57(10), 1457–1469 (2003)
F. Tröltzsch, Optimal Control of Partial Differential Equations: Theory, Methods and Applications. Graduate studies in mathematics, vol. 112 (American Mathematical Society, Providence, 2010)
P. Varignon, Nouvelle mechanique ou statique, 2 vols. (Chez Claude Jombert, A Paris, 1725)
L. von Wolfersdorf, Optimal control for processes governed by mildly nonlinear differential equations of parabolic type I. ZAMM 56, 531–538 (1976)
L. von Wolfersdorf, Optimal control for processes governed by mildly nonlinear differential equations of parabolic type II. ZAMM 57, 11–17 (1977)
E. Zuazua, Some problems and results on the controllability of partial differential equations, in Proceedings of the Second European Conference of Mathematics, Budapest, July 1996. Progress in mathematics (Birkhäuser Verlag, Basel, 1998), pp. 276–311
E. Zuazua, Controllability of partial differential equations and its semi-discrete approximations. Discrete Continuous Dyn. Syst. 8, 469–513 (2002)
Acknowledgements
The authors are grateful to M. Mattmüller for providing us with a copy of Bernoulli’s letter (Univ. Bibl. Basel, Handschriften-Signatur L I a 669, Nr. 50). We further thank Ph. Henry, C. Lubich and E. Hairer for helpful discussions which greatly improved the manuscript. We are also grateful to Armen Sergeev from the Steklov Institute in Moscow for his invaluable help to get the original sources of A.A. Feldbaum, and Peter Kloeden for obtaining the RAND report of Hestenes for us. We thank the Bibliothèque de Genève for granting permission to reproduce photographs from the original sources under catalogue numbers Kc62 (Varignon), Kc110 [33], Kc111 [34], Ka495 [4], Ka368 (Euler’s Methodus E65), Ka459 (Archimedes) and also for Figs. 26, 27, and 28 from [48]. We also thank Tatiana Smirnova-Nagnibeda, Rinat Kashaev, Zdeněk Strakoš and Ivana Gander for their valuable help in translating several texts that originally appeared in Russian. The authors acknowledge support by the European ScienceFoundation, the Swiss National Science Foundation and the Centro Stefano Franscini.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2014 Springer International Publishing Switzerland
About this chapter
Cite this chapter
Gander, M.J., Kwok, F., Wanner, G. (2014). Constrained Optimization: From Lagrangian Mechanics to Optimal Control and PDE Constraints. In: Hoppe, R. (eds) Optimization with PDE Constraints. Lecture Notes in Computational Science and Engineering, vol 101. Springer, Cham. https://doi.org/10.1007/978-3-319-08025-3_5
Download citation
DOI: https://doi.org/10.1007/978-3-319-08025-3_5
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-08024-6
Online ISBN: 978-3-319-08025-3
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)