Sequential Quadratic Programming 1 Lecture 17 ME EN 575 Andrew - - PDF document

sequential quadratic programming 1
SMART_READER_LITE
LIVE PREVIEW

Sequential Quadratic Programming 1 Lecture 17 ME EN 575 Andrew - - PDF document

Sequential Quadratic Programming 1 Lecture 17 ME EN 575 Andrew Ning aning@byu.edu Outline Equality Constrained Quadratic Programming Equality Constrained Sequential Quadratic Programming Quadratic Programming (QP) Equality Constrained


slide-1
SLIDE 1

Sequential Quadratic Programming 1

Lecture 17

ME EN 575 Andrew Ning aning@byu.edu

Outline

Equality Constrained Quadratic Programming Equality Constrained Sequential Quadratic Programming

slide-2
SLIDE 2

Quadratic Programming (QP) Equality Constrained Quadratic Programming

slide-3
SLIDE 3

Pull out a sheet of paper, and try to solve the following analytically. Don’t solve it by variable

  • substitution. Assume that there are lots of

variables and equality constraints and so you couldn’t use that approach. minimize x2 − 8x + y2 − 12y + 48 subject to x + y = 8 Consider a general equality-constrained QP: minimize 1 2xTQx + xTc subject to Ax = b

slide-4
SLIDE 4

We can directly find the global minimum of any equality-constrained QP just by solving a linear system.

∗ As long as Q is positive definite.

slide-5
SLIDE 5

Example: Controls

Linear Dynamical System: xt+1 = Axt + But xt : state at time t ut : control input at time t Define a cost function: J =

N

  • xT

t Qxt + uT t Rut

(cost often represents deviations in the states from target values plus energy expended by control action)

slide-6
SLIDE 6

Example: Aerodynamics

Consider a variation on the wing example I showed last time. Let’s optimize the lift distribution γ.

γ

Equality Constrained Sequential Quadratic Programming

slide-7
SLIDE 7

Now assume we have a general nonlinear equality-constrained optimization problem. minimize f(x) subject to ˆ cj(x) = 0, j = 1, . . . , ˆ m We locally approximate our problem as: minimize 1 2pT∇xxLkp + ∇xLT

k p + Lk

with respect to p subject to ∇xˆ ckp + ˆ ck = 0

slide-8
SLIDE 8

After simplification:  ∇xxLk [∇xˆ ck]T ∇xˆ ck     pk ˆ λk+1   =  −gk −ˆ ck  

SQP Procedure

slide-9
SLIDE 9

An alternative derivation of SQP (now shown here) shows that it is simply an application of Newton’s method (root finding) to the KKT conditions.