HS 2007
151-0563-00
Optimale Regelung
H. P. Geering
2V Mi 13–15 ML F38 ab 26. September
1U Mi 15–16 ML F38 ab 3. Oktober
Skript:
H. P. Geering
Optimal Control with Engineering Applications
Springer, 2007, ISBN 978-3-540-69437-3
Assistenten: Simon Keel, Florian Herzog, Gabriel Dondi
Testatbedingung: 5 von 6 Übungen akzeptiert.
Prüfung: eine halbe Stunde, mündlich
Topics:
• optimal open-loop control (Pontryagin)
• optimal closed-loop control (HJB)
• differential games ( ↔ robust control)
Methods:
• calculus of variations
• geometry
• ordinary and partial differential equations
ProblemA
Problem A:
Optimal control problem with a fixed final state
Find a piecewise continuous control u : [ta , tb ] → Ω ⊆ Rm ,
such that the constraints
x(ta ) = xa
ẋ(t) = f (x(t), u(t), t) t ∈ [ta , tb ]
x(tb ) = xb
are satisfied and such that the cost functional
tb
J(u) = K(tb ) + L(x(t), u(t), t) dt
ta
is minimized;
Subproblem A.1: tb is fixed,
Subproblem A.2: tb is free (tb > ta ).
Remark: ta , xa ∈ Rn , xb ∈ Rn are specified; Ω ⊆ Rm is
time-invariant.
moving target
Problem A.2 with a “moving target”:
Find a piecewise continuous control u : [ta , tb ] → Ω ⊆ Rm ,
such that the constraints
x(ta ) = xa
ẋ(t) = f (x(t), u(t), t) t ∈ [ta , tb ]
x(tb ) = g(tb )
are satisfied and such that the cost functional
tb
J(u) = K(tb ) + L(x(t), u(t), t) dt
ta
is minimized.
Remark: ta , xa ∈ Rn are specified; Ω ⊆ Rm is time-invariant.
ProblemB
Problem B:
Optimal control problem with a partially constrained
final state
Find a piecewise continuous control u : [ta , tb ] → Ω ⊆ Rm ,
such that the constraints
x(ta ) = xa
ẋ(t) = f (x(t), u(t), t) t ∈ [ta , tb ]
x(tb ) ∈ S
are satisfied and such that the cost functional
tb
J(u) = K(x(tb ), tb ) + L(x(t), u(t), t) dt
ta
is minimized;
Subproblem B.1: tb is fixed,
Subproblem B.2: tb is free (tb > ta ).
Remark: ta , xa ∈ Rn , S ⊆ Rn are specified; Ω ⊆ Rm is
time-invariant.
ProblemC
Problem C:
Optimal control problems with a free final state
Find a piecewise continuous control u : [ta , tb ] → Ω ⊆ Rm ,
such that the constraints
x(ta ) = xa
ẋ(t) = f (x(t), u(t), t) t ∈ [ta , tb ]
are satisfied and such that the cost functional
tb
J(u) = K(x(tb ), tb ) + L(x(t), u(t), t) dt
ta
is minimized;
Subproblem C.1: tb is fixed,
Subproblem C.2: tb is free (tb > ta ).
Remark: ta , xa ∈ Rn are specified; Ω ⊆ Rm is time-invariant.
ProblemD
Problem D:
Optimal control problems with state constraints
Find a piecewise continuous control u : [ta , tb ] → Ω ⊆ Rm ,
such that the constraints
x(ta ) = xa
ẋ(t) = f (x(t), u(t), t) t ∈ [ta , tb ]
x(t) ∈ Ωx (t) t ∈ [ta , tb ]
x(tb ) ∈ S
Ωx (t) = {x ∈ Rn | G(x, t) ≤ 0; G : Rn × [ta , tb ] → R}
S = {x ∈ Rn | GS (x) ≤ 0; GS : Rn → R}
are satisfied and such that the cost functional
tb
J(u) = K(x(tb ), tb ) + L(x(t), u(t), t) dt
ta
is minimized;
Subproblem D.1: tb is fixed,
Subproblem D.2: tb is free (tb > ta ).
Remark: ta , xa ∈ Rn are specified, Ω ⊆ Rm is time-invariant.
Static Optimization:
Problem:
Minimize a scalar functons of several variables which are
unconstrainded (S = Rn ) or constrained to a set (S ⊂ Rn ):
f : S ⊆ Rn → R
a) unconstrained:
minimize f (x) for x ∈ Rn
b) constrained:
minimize f (x) for x ∈ S ⊂ Rn
Methods:
• calculus of variations
• Lagrange multiplier method
• geometry
Geometry:
• gradient ∇x f of f (x)
∂f
• hyperplane ∇x f T δx = δx = 0
∂x
• open half space ∇x f T δx > 0
• open half space ∇x f T δx < 0
• tangent cone T (S, xo )
• normal cone T ∗ (S, xo )
• transversality condition
• the Lagrange multiplier vector
• tangent plains of f in Rn+1
• the singular case with λo0 = 0
• geometry of the optimal control problem
at the fixed final time tb
