Generalized mid-point and trapezoidal rules for Lipschitzian RHS
|
|
- Annice Morris
- 5 years ago
- Views:
Transcription
1 Generalized mid-point and trapezoidal rules for Lipschitzian RHS Richard Hasenfelder, Andreas Griewank, Tom Streubel Humboldt Universität zu Berlin 7th International Conference on Algorithmic Differentiation September 016, Oxford Richard Hasenfelder (HU Berlin) Numerical integration via PL models 1 / 36
2 Introduction Introduction Consider initial value problem of the form ẋ(t) = F(x,t), x R n,f : R n R m x(0) = x 0 We assume local Lipschitz continuity of RHS F might be non-smooth (e.g. by using max,min) Most classical methods can drop to first order Richard Hasenfelder (HU Berlin) Numerical integration via PL models / 36
3 Introduction Classical Approach Approximation of Integral: h ˆx ˇx := x(h) x(0) = F(x(t))dt hf 0 ( ) ˇx + ˆx h F(ˇx) + F(ˆx) ˆx ˇx := x(h) x(0) = F(x(t))dt h 0 for midpoint rule and for trapezoidal rule, respectively. These methods have global convergence order for smooth functions. Goal: Modify methods such that they preserve convergence properties even for large number of kinks. Richard Hasenfelder (HU Berlin) Numerical integration via PL models 3 / 36
4 Piecewise Linearization Piecewise Linearization Richard Hasenfelder (HU Berlin) Numerical integration via PL models 4 / 36
5 Piecewise Linearization Piecewise Linearization We propose two methods to approximate locally Lipschitz continuous functions with piecewise affine functions: x ˇx ˆx (a) Tangent mode linearization (b) Secant mode linearization Richard Hasenfelder (HU Berlin) Numerical integration via PL models 5 / 36
6 Piecewise Linearization Piecewise Linearization The proposed approximations have the form F(x) = F( x + x) F( x) + F( x; x) F(x) = F( x + x) 1 [F(ˇx) + F(ˆx)] + F(ˇx, ˆx; x) for tangent mode and for secant mode, respectively, where x = 1 (ˇx + ˆx). Definition A function is called composite piecewise smooth, if it can composed out of some set of smooth elementary functions and the absolute value function, i.e. it can be represented as a directed, acyclic graph with nodes (v i ) i I, where v i+1 = ϕ i+1 (v i ) with ϕ i C 1,1 or abs or a binary operation. = We can construct approximations with AD-like methods. Richard Hasenfelder (HU Berlin) Numerical integration via PL models 6 / 36
7 Propagation Rules I Piecewise Linearization We define the following propagation rules to produce the piecewise linearization in tangent mode Propagation Rules v i = v j ± v k for v i = v j ± v k v i = v j v k + v j v k for v i = v j v k v i = ( v j v k v j v k )/ v k for v i = v j /v k v i = c ij v j for v i = ϕ i (v j ) where we have c ij = ϕ i ( v j) for ϕ i C 1,1 and v i = abs( v i 1 + v i 1 ) v i for ϕ i = abs. Richard Hasenfelder (HU Berlin) Numerical integration via PL models 7 / 36
8 Propagation Rules II Piecewise Linearization For secant mode the basic propagation rules stay the same, Propagation Rules v i = v j ± v k for v i = v j ± v k v i = v j v k + v j v k for v i = v j v k v i = ( v j v k v j v k )/ v k for v i = v j /v k v i = c ij v j for v i = ϕ i (v j ) but here ϕ i ( v j) if ˆx = ˇx c ij = ˆv i ˇv i else ˆv j ˇv j and for ϕ i = abs we have that v i = 1 (ˇv i + ˆv i ) = 1 [abs(ˇv j) + abs(ˆv j )]. Obviously for ˇx = ˆx both rules coincide. Richard Hasenfelder (HU Berlin) Numerical integration via PL models 8 / 36
9 Piecewise Linearization Caculation of Lipschitz Constants In a similar way we can propagate a bound on the Lipschitz constant of F. Calculation of β F v = ϕ(u) = β v = β u ϕ (ů) v = abs(u) = β v = β u v = u + w = β v = β u + β w v = uw = β v = β u ẘ + ů β w v = u w = β v = β u ẘ + ů β w ẘ For this constant it holds: F(x) F( x) β F x x Richard Hasenfelder (HU Berlin) Numerical integration via PL models 9 / 36
10 Piecewise Linearization Caculation of Lipschitz Constants II We can also propagate a constant γ F : Calculation of γ F v = ϕ(u) = γ v = γ u ϕ (ů) + βu ϕ (ů) v = abs(u) = γ v = γ u v = u + w = γ v = γ u + γ w v = uw = γ v = γ u ẘ + β u β w + ů γ w v = u w = γ v = γ u ẘ + ẘ β u β w + ů β w + ů ẘ γ w ẘ 3 γ F is a bound on the Lipschitz constant of the derivative of F (everywhere it exists). Richard Hasenfelder (HU Berlin) Numerical integration via PL models 10 / 36
11 Piecewise Linearization Lipschitz Continuity of the Model The Lipschitz constant β F is also a Lipschitz constant for the piecewise linear model: Lipschitz Continuity x F(x) x F( x) β F x x and ˆxˇx F(x) ˆxˇx F( x) β F x x where: x F(x) F( x) + F( x;x x) and ˆxˇx F(x) 1 (F(ˇx) + F(ˆx)) + F(ˇx, ˆx;x x) Richard Hasenfelder (HU Berlin) Numerical integration via PL models 11 / 36
12 Piecewise Linearization Quality of Approximation We assume that F is composite piecewise Lipschitz continuous differentiable on an open neighbourhood of a closed, convex domain K R n. Then: Tangent mode approximation x,x K : F(x) F( x) F( x;x x) γ F x x. Secant mode approximation ˇx, ˆx,x K : F(x) F F(ˇx, ˆx;x x) γ F x ˇx x ˆx. γ F can be computed as mentioned before. Richard Hasenfelder (HU Berlin) Numerical integration via PL models 1 / 36
13 Generalized methods Generalized methods Richard Hasenfelder (HU Berlin) Numerical integration via PL models 13 / 36
14 Generalized methods Generalized Midpoint Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): h x(h) x(0) = F(x(t))dt 0 Richard Hasenfelder (HU Berlin) Numerical integration via PL models 14 / 36
15 Generalized methods Generalized Midpoint Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): h 1 ˆx ˇx := x(h) x(0) = F(x(t))dt = h 0 1 F ( x ( h + τh)) dτ Richard Hasenfelder (HU Berlin) Numerical integration via PL models 14 / 36
16 Generalized methods Generalized Midpoint Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): h 1 ˆx ˇx := x(h) x(0) = F(x(t))dt = h 0 1 F ( x ( h + τh)) dτ }{{} F(x) F( x)+ F( x; x) Richard Hasenfelder (HU Berlin) Numerical integration via PL models 14 / 36
17 Generalized methods Generalized Midpoint Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): and get h 1 ˆx ˇx := x(h) x(0) = F(x(t))dt = h 0 1 F ( x ( h + τh)) dτ }{{} F(x) F( x)+ F( x; x) 1 ˆx + ˇx ˆx ˇx = h F( x) + F( x;t(ˆx ˇx))dt with x = 1 Richard Hasenfelder (HU Berlin) Numerical integration via PL models 14 / 36
18 Generalized methods Generalized Trapezoidal Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): h 1 ˆx ˇx := x(h) x(0) = F(x(t))dt = h 0 1 F ( x ( h + τh)) dτ Richard Hasenfelder (HU Berlin) Numerical integration via PL models 15 / 36
19 Generalized methods Generalized Trapezoidal Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): h 1 ˆx ˇx := x(h) x(0) = F(x(t))dt = h 0 1 F ( x ( h + τh)) dτ }{{} F(x) F+ F(ˆx,ˇx;t(ˆx ˇx)) Richard Hasenfelder (HU Berlin) Numerical integration via PL models 15 / 36
20 Generalized methods Generalized Trapezoidal Rule Idea for construction: We integrate the ODE ẋ(t) = F(x(t)): h 1 ˆx ˇx := x(h) x(0) = F(x(t))dt = h 0 1 F ( x ( h + τh)) dτ }{{} F(x) F+ F(ˆx,ˇx;t(ˆx ˇx)) and get 1 F(ˆx) + F(ˇx) ˆx ˇx = h F + F(ˆx, ˇx;t(ˆx ˇx))dt with F = 1 Richard Hasenfelder (HU Berlin) Numerical integration via PL models 15 / 36
21 Generalized methods Properties Both rules reduce to familiar rules if F is smooth For F piecewise linear we get 1 ( ) ˆx + ˇx ˆx ˇx = h F + t(ˆx ˇx) dt 1 In piecewise linear case method is an AVF method Trapezoidal rule yields a C 1,1 -interpolant that has correct tangents at ˆx and ˇx In both cases to find new point we have to solve a nonlinear, nonsmooth equation Can be solved with Picard like method for sufficiently small h Richard Hasenfelder (HU Berlin) Numerical integration via PL models 16 / 36
22 Convergence Results Generalized methods The above nonlinear equations can be written in fixed point form and with ˇx = 0 we get [ ] x = h F(x/) F (x/; xt) dt =: hg(x) It can be shown that if we bound h suitably, this equation is contractive. Due to Banach we have a fixed point x h and it can be shown that it holds where x(t) solves the equation x h x(h) = O(h 3 ) ẋ(t) = F(x(t)) with x(0) = 0. Richard Hasenfelder (HU Berlin) Numerical integration via PL models 17 / 36
23 A first numerical result Generalized methods Trapezoida Generalize Impl. Midpoint Trapezoidal Generalized Error in x Number of timesteps n Richard Hasenfelder (HU Berlin) Numerical integration via PL models 18 / 36
24 Generalized methods Theoretical expectations Stable second order convergence without event handling should be as good as or outperform Trapezoidal Rule/Impl. Midpoint Should be able to use (local) Richardson extrapolation to gain higher orders Possible geometric integration properties Richard Hasenfelder (HU Berlin) Numerical integration via PL models 19 / 36
25 Generalized methods Richardson extrapolation With transversality condition = finitely many kinks on smooth parts on kinks global Classical Rule O(h 3 ) O(h ) O(h ) Generalized Rule O(h 3 ) O(h 3 ) O(h ) Class. w/ Romberg O(h 5 ) O(h ) O(h ) Gen. w/ Romberg O(h 5 ) O(h 3 ) O(h 3 ) Richard Hasenfelder (HU Berlin) Numerical integration via PL models 0 / 36
26 Generalized methods Richardson extrapolation Without transversality condition on smooth parts on kinks global Classical Rule O(h 3 ) O(h ) O(h) Generalized Rule O(h 3 ) O(h 3 ) O(h ) Class. w/ Romberg??? Gen. w/ Romberg??? Richard Hasenfelder (HU Berlin) Numerical integration via PL models 1 / 36
27 Error Estimator Error Estimator h x(h) x h = F(x(t))dt 0 h With x(t) p(t) O(h 3 ) we get: 0 h 0 F + F(ˇx, ˆx;t(ˆx ˇx))dt F(x(t)) F((1 t h )ˇx + t h ˆx) dt h + F((1 t h )ˇx + t ˆx) F h + F(ˇx, ˆx;t(ˆx ˇx)) dt 0 h x(h) x h β F p(t) ˇx t (ˆx ˇx) dt + 1 γ h 1 F ˆx ˇx h 0 }{{}}{{} η η 1 Richard Hasenfelder (HU Berlin) Numerical integration via PL models / 36
28 Error Estimator Error Estimator Error Local Error and Estimators Error Trap. Error Est. Gen. Error Est t 1e 8 Richard Hasenfelder (HU Berlin) Numerical integration via PL models 3 / 36
29 Experiments Experiments Richard Hasenfelder (HU Berlin) Numerical integration via PL models 4 / 36
30 Experiments Rolling Stone Rolling Stone ẍ = V (x) with 1 (1 x) if x 1 1 V(x) = (1 + x) if x 1 0 else 1 x if x 1 V (x) = x 1 if x 1 0 else = min(max( x 1,0),1 x) = x 1 x x V(x) V (x) x Figure : RHS of Rolling Stone Richard Hasenfelder (HU Berlin) Numerical integration via PL models 5 / 36
31 x Experiments Rolling Stone Exact solution x(0) = 1, ẋ(0) = 1 x(t) = 1 + sin(t) if t [0,π) 1 (t π) if t [π,π + ) sin( t) if t [π +,π + ) t 3 π if t [π +,π + 4) The total period is π t Figure : Exact solution Richard Hasenfelder (HU Berlin) Numerical integration via PL models 6 / 36
32 Convergence Experiments Rolling Stone Trapezoida Generalize Impl. Midpoint Trapezoidal Generalized Error in x Number of timesteps n Richard Hasenfelder (HU Berlin) Numerical integration via PL models 7 / 36
33 Experiments Rolling Stone Convergence with Romberg extrapolation Convergence Generalized Gen. Romberg Trapezoidal Impl. Midpoint Trap. Romberg Error Number of steps Richard Hasenfelder (HU Berlin) Numerical integration via PL models 8 / 36
34 Experiments Rolling Stone Energy of the System Summed energy loss: V(x) + 1 ẋ Impl. Midpoint Trapezoidal Generalized Mean energy loss Number of timesteps n Richard Hasenfelder (HU Berlin) Numerical integration via PL models 9 / 36
35 Experiments Rolling Stone Energy over Time Energy loss over time P.L. Tapezoidal Rule Trapezoidal Rule Energy t Richard Hasenfelder (HU Berlin) Numerical integration via PL models 30 / 36
36 Experiments Diode-LC-Circuit Problem Formulation ẋ 1 ẏ = F(x) = z ż ( y C sin(ωx) + g 1 (Cz) ) 1 LC L C g(x) V (t) I(t) = z(t) with g 1 (x) = { x α if x 0 x β if x < 0 L = 10 6,C = 10 13,ω = , α =,β = Richard Hasenfelder (HU Berlin) Numerical integration via PL models 31 / 36
37 Experiments Diode-LC-Circuit Computed Solution 3 x 10 4 I(t) 1.8 x Q(t) x x 10 8 Figure : Electric Current Figure : Electric Charge Richard Hasenfelder (HU Berlin) Numerical integration via PL models 3 / 36
38 Experiments Diode-LC-Circuit Convergence Impl. Midpoint Trapezoidal Gen. Midpoint Gen. Trapezoidal Error in x Number of timesteps n Richard Hasenfelder (HU Berlin) Numerical integration via PL models 33 / 36
39 Experiments Diode-LC-Circuit Convergence with Romberg extrapolation 10-9 Convergence Error Number of steps Richard Hasenfelder (HU Berlin) Numerical integration via PL models 34 / 36
40 Experiments Diode-LC-Circuit Convergence of adaptive Solver Convergence Generalized Gen. Romberg Gen. with SSC Trapezoidal Trap. Romberg Trap. with SSC Error Number of steps Richard Hasenfelder (HU Berlin) Numerical integration via PL models 35 / 36
41 Outlook Outlook Improve efficiency of implementation Improve error constant and automatic scaling of error norm Regain energy preservation for extrapolation via restoration of time symmetry Exact integration of PL systems Automatic event handling Generalization to discontinuous case Application to DAEs Application to optimal control problems Application to space discretized PDEs ODE sensitivity using Clark can be wrong, PL isn t (cf. K. Khan) Richard Hasenfelder (HU Berlin) Numerical integration via PL models 36 / 36
On Stable Piecewise Linearization and Generalized Algorithmic Differentiation
On Stable Piecewise Linearization and Generalized Algorithmic Differentiation Andreas Griewank 1 1 Department of Mathematics, Humboldt-University. Berlin December 30, 2012 Keywords Piecewise differentiability,
More informationGeneralized Derivatives Automatic Evaluation & Implications for Algorithms Paul I. Barton, Kamil A. Khan & Harry A. J. Watson
Generalized Derivatives Automatic Evaluation & Implications for Algorithms Paul I. Barton, Kamil A. Khan & Harry A. J. Watson Process Systems Engineering Laboratory Massachusetts Institute of Technology
More informationModified Equations for Variational Integrators
Modified Equations for Variational Integrators Mats Vermeeren Technische Universität Berlin Groningen December 18, 2018 Mats Vermeeren (TU Berlin) Modified equations for variational integrators December
More informationz x = f x (x, y, a, b), z y = f y (x, y, a, b). F(x, y, z, z x, z y ) = 0. This is a PDE for the unknown function of two independent variables.
Chapter 2 First order PDE 2.1 How and Why First order PDE appear? 2.1.1 Physical origins Conservation laws form one of the two fundamental parts of any mathematical model of Continuum Mechanics. These
More information2tdt 1 y = t2 + C y = which implies C = 1 and the solution is y = 1
Lectures - Week 11 General First Order ODEs & Numerical Methods for IVPs In general, nonlinear problems are much more difficult to solve than linear ones. Unfortunately many phenomena exhibit nonlinear
More informationPreliminary Examination, Numerical Analysis, August 2016
Preliminary Examination, Numerical Analysis, August 2016 Instructions: This exam is closed books and notes. The time allowed is three hours and you need to work on any three out of questions 1-4 and any
More informationCHAPTER 10: Numerical Methods for DAEs
CHAPTER 10: Numerical Methods for DAEs Numerical approaches for the solution of DAEs divide roughly into two classes: 1. direct discretization 2. reformulation (index reduction) plus discretization Direct
More informationAD for PS Functions and Relations to Generalized Derivative Concepts
AD for PS Functions and Relations to Generalized Derivative Concepts Andrea Walther 1 and Andreas Griewank 2 1 Institut für Mathematik, Universität Paderborn 2 School of Math. Sciences and Information
More informationNumerical Solutions to Partial Differential Equations
Numerical Solutions to Partial Differential Equations Zhiping Li LMAM and School of Mathematical Sciences Peking University Discretization of Boundary Conditions Discretization of Boundary Conditions On
More informationCS 450 Numerical Analysis. Chapter 8: Numerical Integration and Differentiation
Lecture slides based on the textbook Scientific Computing: An Introductory Survey by Michael T. Heath, copyright c 2018 by the Society for Industrial and Applied Mathematics. http://www.siam.org/books/cl80
More informationObstacle problems and isotonicity
Obstacle problems and isotonicity Thomas I. Seidman Revised version for NA-TMA: NA-D-06-00007R1+ [June 6, 2006] Abstract For variational inequalities of an abstract obstacle type, a comparison principle
More informationScientific Computing: Numerical Integration
Scientific Computing: Numerical Integration Aleksandar Donev Courant Institute, NYU 1 donev@courant.nyu.edu 1 Course MATH-GA.2043 or CSCI-GA.2112, Fall 2015 Nov 5th, 2015 A. Donev (Courant Institute) Lecture
More informationVerification of analog and mixed-signal circuits using hybrid systems techniques
FMCAD, November 2004, Austin Verification of analog and mixed-signal circuits using hybrid systems techniques Thao Dang, Alexandre Donze, Oded Maler VERIMAG Grenoble, France Plan 1. Introduction 2. Verification
More informationKrylov Subspace Methods for Nonlinear Model Reduction
MAX PLANCK INSTITUT Conference in honour of Nancy Nichols 70th birthday Reading, 2 3 July 2012 Krylov Subspace Methods for Nonlinear Model Reduction Peter Benner and Tobias Breiten Max Planck Institute
More informationFINITE-DIFFERENCE APPROXIMATIONS AND OPTIMAL CONTROL OF THE SWEEPING PROCESS. BORIS MORDUKHOVICH Wayne State University, USA
FINITE-DIFFERENCE APPROXIMATIONS AND OPTIMAL CONTROL OF THE SWEEPING PROCESS BORIS MORDUKHOVICH Wayne State University, USA International Workshop Optimization without Borders Tribute to Yurii Nesterov
More informationLecture Note 7: Switching Stabilization via Control-Lyapunov Function
ECE7850: Hybrid Systems:Theory and Applications Lecture Note 7: Switching Stabilization via Control-Lyapunov Function Wei Zhang Assistant Professor Department of Electrical and Computer Engineering Ohio
More informationThe first order quasi-linear PDEs
Chapter 2 The first order quasi-linear PDEs The first order quasi-linear PDEs have the following general form: F (x, u, Du) = 0, (2.1) where x = (x 1, x 2,, x 3 ) R n, u = u(x), Du is the gradient of u.
More informationGeneralized Pattern Search Algorithms : unconstrained and constrained cases
IMA workshop Optimization in simulation based models Generalized Pattern Search Algorithms : unconstrained and constrained cases Mark A. Abramson Air Force Institute of Technology Charles Audet École Polytechnique
More informationStability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games
Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games Alberto Bressan ) and Khai T. Nguyen ) *) Department of Mathematics, Penn State University **) Department of Mathematics,
More informationJim Lambers MAT 460/560 Fall Semester Practice Final Exam
Jim Lambers MAT 460/560 Fall Semester 2009-10 Practice Final Exam 1. Let f(x) = sin 2x + cos 2x. (a) Write down the 2nd Taylor polynomial P 2 (x) of f(x) centered around x 0 = 0. (b) Write down the corresponding
More informationUniversity of Houston, Department of Mathematics Numerical Analysis, Fall 2005
3 Numerical Solution of Nonlinear Equations and Systems 3.1 Fixed point iteration Reamrk 3.1 Problem Given a function F : lr n lr n, compute x lr n such that ( ) F(x ) = 0. In this chapter, we consider
More informationCS 257: Numerical Methods
CS 57: Numerical Methods Final Exam Study Guide Version 1.00 Created by Charles Feng http://www.fenguin.net CS 57: Numerical Methods Final Exam Study Guide 1 Contents 1 Introductory Matter 3 1.1 Calculus
More informationyou expect to encounter difficulties when trying to solve A x = b? 4. A composite quadrature rule has error associated with it in the following form
Qualifying exam for numerical analysis (Spring 2017) Show your work for full credit. If you are unable to solve some part, attempt the subsequent parts. 1. Consider the following finite difference: f (0)
More informationNonlinear Control Lecture # 14 Input-Output Stability. Nonlinear Control
Nonlinear Control Lecture # 14 Input-Output Stability L Stability Input-Output Models: y = Hu u(t) is a piecewise continuous function of t and belongs to a linear space of signals The space of bounded
More informationPartial Differential Equations
Part II Partial Differential Equations Year 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2015 Paper 4, Section II 29E Partial Differential Equations 72 (a) Show that the Cauchy problem for u(x,
More informationSECOND ORDER TIME DISCONTINUOUS GALERKIN METHOD FOR NONLINEAR CONVECTION-DIFFUSION PROBLEMS
Proceedings of ALGORITMY 2009 pp. 1 10 SECOND ORDER TIME DISCONTINUOUS GALERKIN METHOD FOR NONLINEAR CONVECTION-DIFFUSION PROBLEMS MILOSLAV VLASÁK Abstract. We deal with a numerical solution of a scalar
More informationC.-H. Lamarque. University of Lyon/ENTPE/LGCB & LTDS UMR CNRS 5513
Nonlinear Dynamics of Smooth and Non-Smooth Systems with Application to Passive Controls 3rd Sperlonga Summer School on Mechanics and Engineering Sciences on Dynamics, Stability and Control of Flexible
More informationNumerical Algorithms for ODEs/DAEs (Transient Analysis)
Numerical Algorithms for ODEs/DAEs (Transient Analysis) Slide 1 Solving Differential Equation Systems d q ( x(t)) + f (x(t)) + b(t) = 0 dt DAEs: many types of solutions useful DC steady state: state no
More information1. Method 1: bisection. The bisection methods starts from two points a 0 and b 0 such that
Chapter 4 Nonlinear equations 4.1 Root finding Consider the problem of solving any nonlinear relation g(x) = h(x) in the real variable x. We rephrase this problem as one of finding the zero (root) of a
More informationJoint work with Nguyen Hoang (Univ. Concepción, Chile) Padova, Italy, May 2018
EXTENDED EULER-LAGRANGE AND HAMILTONIAN CONDITIONS IN OPTIMAL CONTROL OF SWEEPING PROCESSES WITH CONTROLLED MOVING SETS BORIS MORDUKHOVICH Wayne State University Talk given at the conference Optimization,
More information( ) and then eliminate t to get y(x) or x(y). Not that the
2. First-order linear equations We want the solution (x,y) of () ax, ( y) x + bx, ( y) y = 0 ( ) and b( x,y) are given functions. If a and b are constants, then =F(bx-ay) where ax, y where F is an arbitrary
More informationThe Implicit and Inverse Function Theorems Notes to supplement Chapter 13.
The Implicit and Inverse Function Theorems Notes to supplement Chapter 13. Remark: These notes are still in draft form. Examples will be added to Section 5. If you see any errors, please let me know. 1.
More informationHybrid Control and Switched Systems. Lecture #9 Analysis tools for hybrid systems: Impact maps
Hybrid Control and Switched Systems Lecture #9 Analysis tools for hybrid systems: Impact maps João P. Hespanha University of California at Santa Barbara Summary Analysis tools for hybrid systems Impact
More informationShiqian Ma, MAT-258A: Numerical Optimization 1. Chapter 4. Subgradient
Shiqian Ma, MAT-258A: Numerical Optimization 1 Chapter 4 Subgradient Shiqian Ma, MAT-258A: Numerical Optimization 2 4.1. Subgradients definition subgradient calculus duality and optimality conditions Shiqian
More informationOn rigorous integration of piece-wise linear systems
On rigorous integration of piece-wise linear systems Zbigniew Galias Department of Electrical Engineering AGH University of Science and Technology 06.2011, Trieste, Italy On rigorous integration of PWL
More informationSECTION C: CONTINUOUS OPTIMISATION LECTURE 9: FIRST ORDER OPTIMALITY CONDITIONS FOR CONSTRAINED NONLINEAR PROGRAMMING
Nf SECTION C: CONTINUOUS OPTIMISATION LECTURE 9: FIRST ORDER OPTIMALITY CONDITIONS FOR CONSTRAINED NONLINEAR PROGRAMMING f(x R m g HONOUR SCHOOL OF MATHEMATICS, OXFORD UNIVERSITY HILARY TERM 5, DR RAPHAEL
More informationAn introduction to Mathematical Theory of Control
An introduction to Mathematical Theory of Control Vasile Staicu University of Aveiro UNICA, May 2018 Vasile Staicu (University of Aveiro) An introduction to Mathematical Theory of Control UNICA, May 2018
More informationNonlinear Control. Nonlinear Control Lecture # 2 Stability of Equilibrium Points
Nonlinear Control Lecture # 2 Stability of Equilibrium Points Basic Concepts ẋ = f(x) f is locally Lipschitz over a domain D R n Suppose x D is an equilibrium point; that is, f( x) = 0 Characterize and
More informationLecture 19: Heat conduction with distributed sources/sinks
Introductory lecture notes on Partial Differential Equations - c Anthony Peirce. Not to be copied, used, or revised without explicit written permission from the copyright owner. 1 ecture 19: Heat conduction
More informationLMI Methods in Optimal and Robust Control
LMI Methods in Optimal and Robust Control Matthew M. Peet Arizona State University Lecture 15: Nonlinear Systems and Lyapunov Functions Overview Our next goal is to extend LMI s and optimization to nonlinear
More informationSec. 1.1: Basics of Vectors
Sec. 1.1: Basics of Vectors Notation for Euclidean space R n : all points (x 1, x 2,..., x n ) in n-dimensional space. Examples: 1. R 1 : all points on the real number line. 2. R 2 : all points (x 1, x
More informationExtrapolation in Numerical Integration. Romberg Integration
Extrapolation in Numerical Integration Romberg Integration Matthew Battaglia Joshua Berge Sara Case Yoobin Ji Jimu Ryoo Noah Wichrowski Introduction Extrapolation: the process of estimating beyond the
More informationImplicit-explicit exponential integrators
Implicit-explicit exponential integrators Bawfeh Kingsley Kometa joint work with Elena Celledoni MaGIC 2011 Finse, March 1-4 1 Introduction Motivation 2 semi-lagrangian Runge-Kutta exponential integrators
More informationOn second order sufficient optimality conditions for quasilinear elliptic boundary control problems
On second order sufficient optimality conditions for quasilinear elliptic boundary control problems Vili Dhamo Technische Universität Berlin Joint work with Eduardo Casas Workshop on PDE Constrained Optimization
More informationNow I switch to nonlinear systems. In this chapter the main object of study will be
Chapter 4 Stability 4.1 Autonomous systems Now I switch to nonlinear systems. In this chapter the main object of study will be ẋ = f(x), x(t) X R k, f : X R k, (4.1) where f is supposed to be locally Lipschitz
More informationDifferential Equations
Differential Equations Overview of differential equation! Initial value problem! Explicit numeric methods! Implicit numeric methods! Modular implementation Physics-based simulation An algorithm that
More informationParallel-in-time integrators for Hamiltonian systems
Parallel-in-time integrators for Hamiltonian systems Claude Le Bris ENPC and INRIA Visiting Professor, The University of Chicago joint work with X. Dai (Paris 6 and Chinese Academy of Sciences), F. Legoll
More informationNonlinear Control Systems
Nonlinear Control Systems António Pedro Aguiar pedro@isr.ist.utl.pt 3. Fundamental properties IST-DEEC PhD Course http://users.isr.ist.utl.pt/%7epedro/ncs2012/ 2012 1 Example Consider the system ẋ = f
More informationarxiv: v2 [cs.sy] 7 Apr 2017
arxiv:6.58v [cs.sy] 7 Apr 7 Observer design for piecewise smooth and switched systems via contraction theory Davide Fiore Marco Coraggio Mario di Bernardo, Department of Electrical Engineering and Information
More informationDesign of optimal RF pulses for NMR as a discrete-valued control problem
Design of optimal RF pulses for NMR as a discrete-valued control problem Christian Clason Faculty of Mathematics, Universität Duisburg-Essen joint work with Carla Tameling (Göttingen) and Benedikt Wirth
More informationHigh-Order Representation of Poincaré Maps
High-Order Representation of Poincaré Maps Johannes Grote, Martin Berz, and Kyoko Makino Department of Physics and Astronomy, Michigan State University, East Lansing, MI, USA. {grotejoh,berz,makino}@msu.edu
More informationUsing Simulink to analyze 2 degrees of freedom system
Using Simulink to analyze 2 degrees of freedom system Nasser M. Abbasi Spring 29 page compiled on June 29, 25 at 4:2pm Abstract A two degrees of freedom system consisting of two masses connected by springs
More informationOn Nonlinear Dirichlet Neumann Algorithms for Jumping Nonlinearities
On Nonlinear Dirichlet Neumann Algorithms for Jumping Nonlinearities Heiko Berninger, Ralf Kornhuber, and Oliver Sander FU Berlin, FB Mathematik und Informatik (http://www.math.fu-berlin.de/rd/we-02/numerik/)
More informationConvexity of the Reachable Set of Nonlinear Systems under L 2 Bounded Controls
1 1 Convexity of the Reachable Set of Nonlinear Systems under L 2 Bounded Controls B.T.Polyak Institute for Control Science, Moscow, Russia e-mail boris@ipu.rssi.ru Abstract Recently [1, 2] the new convexity
More informationMAT 311 Midterm #1 Show your work! 1. The existence and uniqueness theorem says that, given a point (x 0, y 0 ) the ODE. y = (1 x 2 y 2 ) 1/3
MAT 3 Midterm # Show your work!. The existence and uniqueness theorem says that, given a point (x 0, y 0 ) the ODE y = ( x 2 y 2 ) /3 has a unique (local) solution with initial condition y(x 0 ) = y 0
More informationMATH 2053 Calculus I Review for the Final Exam
MATH 05 Calculus I Review for the Final Exam (x+ x) 9 x 9 1. Find the limit: lim x 0. x. Find the limit: lim x + x x (x ).. Find lim x (x 5) = L, find such that f(x) L < 0.01 whenever 0 < x
More informationNumerical Integration of Functions
Numerical Integration of Functions Berlin Chen Department of Computer Science & Information Engineering National Taiwan Normal University Reference: 1. Applied Numerical Methods with MATLAB for Engineers,
More informationThe Direct Transcription Method For Optimal Control. Part 2: Optimal Control
The Direct Transcription Method For Optimal Control Part 2: Optimal Control John T Betts Partner, Applied Mathematical Analysis, LLC 1 Fundamental Principle of Transcription Methods Transcription Method
More informationStructurally Stable Singularities for a Nonlinear Wave Equation
Structurally Stable Singularities for a Nonlinear Wave Equation Alberto Bressan, Tao Huang, and Fang Yu Department of Mathematics, Penn State University University Park, Pa. 1682, U.S.A. e-mails: bressan@math.psu.edu,
More informationFundamental Solution
Fundamental Solution onsider the following generic equation: Lu(X) = f(x). (1) Here X = (r, t) is the space-time coordinate (if either space or time coordinate is absent, then X t, or X r, respectively);
More informationA note on accurate and efficient higher order Galerkin time stepping schemes for the nonstationary Stokes equations
A note on accurate and efficient higher order Galerkin time stepping schemes for the nonstationary Stokes equations S. Hussain, F. Schieweck, S. Turek Abstract In this note, we extend our recent work for
More informationHow to solve quasi linear first order PDE. v(u, x) Du(x) = f(u, x) on U IR n, (1)
How to solve quasi linear first order PDE A quasi linear PDE is an equation of the form subject to the initial condition v(u, x) Du(x) = f(u, x) on U IR n, (1) u = g on Γ, (2) where Γ is a hypersurface
More informationA Second Order Discontinuous Galerkin Fast Sweeping Method for Eikonal Equations
A Second Order Discontinuous Galerkin Fast Sweeping Method for Eikonal Equations Fengyan Li, Chi-Wang Shu, Yong-Tao Zhang 3, and Hongkai Zhao 4 ABSTRACT In this paper, we construct a second order fast
More informationNonlinear equations. Norms for R n. Convergence orders for iterative methods
Nonlinear equations Norms for R n Assume that X is a vector space. A norm is a mapping X R with x such that for all x, y X, α R x = = x = αx = α x x + y x + y We define the following norms on the vector
More informationNested Differentiation and Symmetric Hessian Graphs with ADTAGEO
Nested Differentiation and Symmetric Hessian Graphs with ADTAGEO ADjoints and TAngents by Graph Elimination Ordering Jan Riehme Andreas Griewank Institute for Applied Mathematics Humboldt Universität zu
More informationFast evaluation of mixed derivatives and calculation of optimal weights for integration. Hernan Leovey
Fast evaluation of mixed derivatives and calculation of optimal weights for integration Humboldt Universität zu Berlin 02.14.2012 MCQMC2012 Tenth International Conference on Monte Carlo and Quasi Monte
More informationAffine Arithmetic: Concepts and Applications
SCAN 2002 Affine Arithmetic: Concepts and Applications Luiz Henrique de Figueiredo (IMPA) Jorge Stolfi (UNICAMP) Outline What is affine arithmetic? The dependency problem in interval arithmetic Main concepts
More information2.152 Course Notes Contraction Analysis MIT, 2005
2.152 Course Notes Contraction Analysis MIT, 2005 Jean-Jacques Slotine Contraction Theory ẋ = f(x, t) If Θ(x, t) such that, uniformly x, t 0, F = ( Θ + Θ f x )Θ 1 < 0 Θ(x, t) T Θ(x, t) > 0 then all solutions
More informationNEWTON S METHOD IN THE CONTEXT OF GRADIENTS
Electronic Journal of Differential Equations, Vol. 2007(2007), No. 124, pp. 1 13. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu (login: ftp) NEWTON
More informationAxioms of Adaptivity (AoA) in Lecture 3 (sufficient for optimal convergence rates)
Axioms of Adaptivity (AoA) in Lecture 3 (sufficient for optimal convergence rates) Carsten Carstensen Humboldt-Universität zu Berlin 2018 International Graduate Summer School on Frontiers of Applied and
More informationTaylor Series and stationary points
Chapter 5 Taylor Series and stationary points 5.1 Taylor Series The surface z = f(x, y) and its derivatives can give a series approximation for f(x, y) about some point (x 0, y 0 ) as illustrated in Figure
More informationDefect-based a-posteriori error estimation for implicit ODEs and DAEs
1 / 24 Defect-based a-posteriori error estimation for implicit ODEs and DAEs W. Auzinger Institute for Analysis and Scientific Computing Vienna University of Technology Workshop on Innovative Integrators
More informationPartitioned Methods for Multifield Problems
C Partitioned Methods for Multifield Problems Joachim Rang, 6.7.2016 6.7.2016 Joachim Rang Partitioned Methods for Multifield Problems Seite 1 C One-dimensional piston problem fixed wall Fluid flexible
More informationOptimization and Optimal Control in Banach Spaces
Optimization and Optimal Control in Banach Spaces Bernhard Schmitzer October 19, 2017 1 Convex non-smooth optimization with proximal operators Remark 1.1 (Motivation). Convex optimization: easier to solve,
More informationNumerical Integration
Numerical Integration Sanzheng Qiao Department of Computing and Software McMaster University February, 2014 Outline 1 Introduction 2 Rectangle Rule 3 Trapezoid Rule 4 Error Estimates 5 Simpson s Rule 6
More information2.1 Dynamical systems, phase flows, and differential equations
Chapter 2 Fundamental theorems 2.1 Dynamical systems, phase flows, and differential equations A dynamical system is a mathematical formalization of an evolutionary deterministic process. An evolutionary
More informationDifferential Equation Types. Moritz Diehl
Differential Equation Types Moritz Diehl Overview Ordinary Differential Equations (ODE) Differential Algebraic Equations (DAE) Partial Differential Equations (PDE) Delay Differential Equations (DDE) Ordinary
More informationIntegration, differentiation, and root finding. Phys 420/580 Lecture 7
Integration, differentiation, and root finding Phys 420/580 Lecture 7 Numerical integration Compute an approximation to the definite integral I = b Find area under the curve in the interval Trapezoid Rule:
More informationChapter 1. Optimality Conditions: Unconstrained Optimization. 1.1 Differentiable Problems
Chapter 1 Optimality Conditions: Unconstrained Optimization 1.1 Differentiable Problems Consider the problem of minimizing the function f : R n R where f is twice continuously differentiable on R n : P
More informationHalf of Final Exam Name: Practice Problems October 28, 2014
Math 54. Treibergs Half of Final Exam Name: Practice Problems October 28, 24 Half of the final will be over material since the last midterm exam, such as the practice problems given here. The other half
More informationRelevant self-assessment exercises: [LIST SELF-ASSESSMENT EXERCISES HERE]
Chapter 6 Finite Volume Methods In the previous chapter we have discussed finite difference methods for the discretization of PDEs. In developing finite difference methods we started from the differential
More informationPractice Problems for Final Exam
Math 1280 Spring 2016 Practice Problems for Final Exam Part 2 (Sections 6.6, 6.7, 6.8, and chapter 7) S o l u t i o n s 1. Show that the given system has a nonlinear center at the origin. ẋ = 9y 5y 5,
More informationChap. 3. Controlled Systems, Controllability
Chap. 3. Controlled Systems, Controllability 1. Controllability of Linear Systems 1.1. Kalman s Criterion Consider the linear system ẋ = Ax + Bu where x R n : state vector and u R m : input vector. A :
More informationb) The system of ODE s d x = v(x) in U. (2) dt
How to solve linear and quasilinear first order partial differential equations This text contains a sketch about how to solve linear and quasilinear first order PDEs and should prepare you for the general
More informationModeling & Simulation 2018 Lecture 12. Simulations
Modeling & Simulation 2018 Lecture 12. Simulations Claudio Altafini Automatic Control, ISY Linköping University, Sweden Summary of lecture 7-11 1 / 32 Models of complex systems physical interconnections,
More informationMCE693/793: Analysis and Control of Nonlinear Systems
MCE693/793: Analysis and Control of Nonlinear Systems Systems of Differential Equations Phase Plane Analysis Hanz Richter Mechanical Engineering Department Cleveland State University Systems of Nonlinear
More informationRomberg Integration and Gaussian Quadrature
Romberg Integration and Gaussian Quadrature P. Sam Johnson October 17, 014 P. Sam Johnson (NITK) Romberg Integration and Gaussian Quadrature October 17, 014 1 / 19 Overview We discuss two methods for integration.
More informationTHE BANG-BANG PRINCIPLE FOR THE GOURSAT-DARBOUX PROBLEM*
THE BANG-BANG PRINCIPLE FOR THE GOURSAT-DARBOUX PROBLEM* DARIUSZ IDCZAK Abstract. In the paper, the bang-bang principle for a control system connected with a system of linear nonautonomous partial differential
More informationAB.Q103.NOTES: Chapter 2.4, 3.1, 3.2 LESSON 1. Discovering the derivative at x = a: Slopes of secants and tangents to a curve
AB.Q103.NOTES: Chapter 2.4, 3.1, 3.2 LESSON 1 Discovering the derivative at x = a: Slopes of secants and tangents to a curve 1 1. Instantaneous rate of change versus average rate of change Equation of
More informationFrom Completing the Squares and Orthogonal Projection to Finite Element Methods
From Completing the Squares and Orthogonal Projection to Finite Element Methods Mo MU Background In scientific computing, it is important to start with an appropriate model in order to design effective
More informationx n+1 = x n f(x n) f (x n ), n 0.
1. Nonlinear Equations Given scalar equation, f(x) = 0, (a) Describe I) Newtons Method, II) Secant Method for approximating the solution. (b) State sufficient conditions for Newton and Secant to converge.
More informationNumerical Methods. King Saud University
Numerical Methods King Saud University Aims In this lecture, we will... find the approximate solutions of derivative (first- and second-order) and antiderivative (definite integral only). Numerical Differentiation
More informationFIXED POINT ITERATIONS
FIXED POINT ITERATIONS MARKUS GRASMAIR 1. Fixed Point Iteration for Non-linear Equations Our goal is the solution of an equation (1) F (x) = 0, where F : R n R n is a continuous vector valued mapping in
More informationOutline. 1 Numerical Integration. 2 Numerical Differentiation. 3 Richardson Extrapolation
Outline Numerical Integration Numerical Differentiation Numerical Integration Numerical Differentiation 3 Michael T. Heath Scientific Computing / 6 Main Ideas Quadrature based on polynomial interpolation:
More information1 Discretizing BVP with Finite Element Methods.
1 Discretizing BVP with Finite Element Methods In this section, we will discuss a process for solving boundary value problems numerically, the Finite Element Method (FEM) We note that such method is a
More informationLecture 20: Bezier Curves & Splines
Lecture 20: Bezier Curves & Splines December 6, 2016 12/6/16 CSU CS410 Bruce Draper & J. Ross Beveridge 1 Review: The Pen Metaphore Think of putting a pen to paper Pen position described by time t Seeing
More informationSUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS
SUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS A. RÖSCH AND R. SIMON Abstract. An optimal control problem for an elliptic equation
More information(f(x) P 3 (x)) dx. (a) The Lagrange formula for the error is given by
1. QUESTION (a) Given a nth degree Taylor polynomial P n (x) of a function f(x), expanded about x = x 0, write down the Lagrange formula for the truncation error, carefully defining all its elements. How
More informationConvergent systems: analysis and synthesis
Convergent systems: analysis and synthesis Alexey Pavlov, Nathan van de Wouw, and Henk Nijmeijer Eindhoven University of Technology, Department of Mechanical Engineering, P.O.Box. 513, 5600 MB, Eindhoven,
More informationMath 212-Lecture 8. The chain rule with one independent variable
Math 212-Lecture 8 137: The multivariable chain rule The chain rule with one independent variable w = f(x, y) If the particle is moving along a curve x = x(t), y = y(t), then the values that the particle
More information