Differential Equation Axiomatization The Impressive Power of - - PowerPoint PPT Presentation

differential equation axiomatization
SMART_READER_LITE
LIVE PREVIEW

Differential Equation Axiomatization The Impressive Power of - - PowerPoint PPT Presentation

Differential Equation Axiomatization The Impressive Power of Differential Ghosts Andr e Platzer Yong Kiam Tan 0.5 0.4 0.3 0.2 1.0 0.1 0.8 0.6 0.4 0.2 Andr e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization


slide-1
SLIDE 1

Differential Equation Axiomatization

The Impressive Power of Differential Ghosts Andr´ e Platzer Yong Kiam Tan

0.2 0.4 0.6 0.8 1.0

0.1 0.2 0.3 0.4 0.5

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 1 / 18

slide-2
SLIDE 2

Outline

1

Differential Dynamic Logic

2

Proofs for Differential Equations Differential Invariants / Cuts / Ghosts

3

Completeness for Differential Equation Invariants Darboux are Differential Ghosts Derived Semialgebraic Invariants Real Induction Derived Local Progress Completeness for Invariants

4

Summary

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 1 / 18

slide-3
SLIDE 3

Hybrid Systems: Example Robot Control

Challenge (Hybrid Systems)

Fixed law describing state evolution with both Discrete dynamics (control decisions) Continuous dynamics (differential equations)

1 2 3 4 5 6 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

1 2 3 4 5 t 2 1 1 2a 1 2 3 4 5 t 0.0 0.5 1.0 1.5 2.0 2.5 3.0v

m

1 2 3 4 5 t 1 2 3 4 5 6 7x

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 2 / 18

slide-4
SLIDE 4

Contributions: Differential Equation Axiomatization

Classical approach: 1 Given ODE 2 Solve ODE 3 Analyze solution Descriptive power of ODEs: ODE much easier than its solution Analyzing ODEs via their solutions undoes their descriptive power! describe ODE describe solution analyze ODE analyze solution

  • Poincar´

e 1881

1 Now: Logical foundations of differential equation invariants 2 Identify axioms for differential equations 3 Completeness for differential equation invariants 4 Uniformly substitutable axioms, not infinite axiom schemata 5 Decide invariance by proof Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 3 / 18

slide-5
SLIDE 5

Outline

1

Differential Dynamic Logic

2

Proofs for Differential Equations Differential Invariants / Cuts / Ghosts

3

Completeness for Differential Equation Invariants Darboux are Differential Ghosts Derived Semialgebraic Invariants Real Induction Derived Local Progress Completeness for Invariants

4

Summary

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 3 / 18

slide-6
SLIDE 6

Hybrid Systems = Differential Equations + Discrete

Concept (Differential Dynamic Logic) (JAR’08,LICS’12)

u2 ≤ v2+9 2 → [u′ = −v+u 4(1−u2−v2), v′ = u+v 4(1−u2−v2)] u2 ≤ v2+9 2 u2+v2 = 1 → [u′ = −v+u 4(1−u2−v2), v′ = u+v 4(1−u2−v2)] u2+v2 = 1

  • 2

2

u

  • 1

1

v

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 4 / 18

slide-7
SLIDE 7

Outline

1

Differential Dynamic Logic

2

Proofs for Differential Equations Differential Invariants / Cuts / Ghosts

3

Completeness for Differential Equation Invariants Darboux are Differential Ghosts Derived Semialgebraic Invariants Real Induction Derived Local Progress Completeness for Invariants

4

Summary

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 4 / 18

slide-8
SLIDE 8

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-9
SLIDE 9

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-10
SLIDE 10

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-11
SLIDE 11

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-12
SLIDE 12

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-13
SLIDE 13

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-14
SLIDE 14

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-15
SLIDE 15

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x) y′ = g(x, y)

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-16
SLIDE 16

Differential Invariants for Differential Equations

Differential Invariant Differential Cut Differential Ghost

t x x′ = f (x) y′ = g(x, y) inv

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 5 / 18

slide-17
SLIDE 17

Differential Invariants for Differential Equations

Differential Invariant Q ⊢ [x′ := f (x)](P)′ P ⊢ [x′ = f (x) & Q]P

x w u r x′ = f(x) & Q P w Q

Differential Cut P ⊢ [x′ = f (x) & Q]C P ⊢ [x′ = f (x) & Q∧C]P P ⊢ [x′ = f (x) & Q]P

x Q w u r x′ = f(x) & Q C w Q

Differential Ghost P ↔ ∃y G G ⊢ [x′ = f (x), y′ = g(x, y) & Q]G P ⊢ [x′ = f (x) & Q]P

x Q w u r x′ = f(x) & Q

deductive power adds DI ≺ DC ≺ DG JLogComput’10,LMCS’12, LICS’12,JAR’17,LICS’18

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 6 / 18

slide-18
SLIDE 18

Differential Invariants for Differential Equations

Differential Invariant Q ⊢ [x′ := f (x)](P)′ P ⊢ [x′ = f (x) & Q]P

x w u r x′ = f(x) & Q P w Q

Differential Cut P ⊢ [x′ = f (x) & Q]C P ⊢ [x′ = f (x) & Q∧C]P P ⊢ [x′ = f (x) & Q]P

x Q w u r x′ = f(x) & Q C w Q

Differential Ghost P ↔ ∃y G G ⊢ [x′ = f (x), y′ = g(x, y) & Q]G P ⊢ [x′ = f (x) & Q]P

x Q w u r x′ = f(x) & Q

if new y′ = g(x, y) has long enough solution JLogComput’10,LMCS’12, LICS’12,JAR’17,LICS’18

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 6 / 18

slide-19
SLIDE 19

Outline

1

Differential Dynamic Logic

2

Proofs for Differential Equations Differential Invariants / Cuts / Ghosts

3

Completeness for Differential Equation Invariants Darboux are Differential Ghosts Derived Semialgebraic Invariants Real Induction Derived Local Progress Completeness for Invariants

4

Summary

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 6 / 18

slide-20
SLIDE 20

Differential Equation Axiomatization

Theorem (Algebraic Completeness) (LICS’18)

dL calculus is a sound & complete axiomatization of algebraic invariants of polynomial differential equations. They are decidable by DI,DC,DG

Theorem (Semialgebraic Completeness) (LICS’18)

dL calculus with RI is a sound & complete axiomatization of semialgebraic invariants of differential equations. They are decidable in dL

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 7 / 18

slide-21
SLIDE 21

ODE Axiomatization: Derived Darboux Rules

Darboux equalities are DG Q ⊢ p′ = gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 (g ∈ R[x]) Definable p′ for Lie-derivative w.r.t. ODE Gaston Darboux 1878

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-22
SLIDE 22

ODE Axiomatization: Derived Darboux Rules

Darboux equalities are DG Q ⊢ p′ = gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 (g ∈ R[x]) ⊢ 2uu′+2vv′ = 2(u2+v2)(u2+v2−1) .. ⊢ [u′ = −v−u+u3+uv2 v′ = u−v+u2v+v3] u2+v2−1=0 Gaston Darboux 1878

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-23
SLIDE 23

ODE Axiomatization: Derived Darboux Rules

Darboux equalities are DG Q ⊢ p′ = gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 (g ∈ R[x])

Proof Idea.

1 DG counterweight y′ = −gy to reduce p = 0 to py = 0 ∧ y = 0. 2 DG counter-counterweight z′ = gz to reduce y = 0 to yz = 1. 3 py = 0 and yz = 1 are now differential invariants by construction. Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-24
SLIDE 24

ODE Axiomatization: Derived Darboux Rules

Darboux inequalities are DG Q ⊢ p′ ≥ gp p 0 ⊢ [x′ = f (x) & Q]p 0 (g ∈ R[x]) Thomas Gr¨

  • nwall 1919

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-25
SLIDE 25

ODE Axiomatization: Derived Darboux Rules

Darboux inequalities are DG Q ⊢ p′ ≥ gp p 0 ⊢ [x′ = f (x) & Q]p 0 (g ∈ R[x])

Proof Idea.

1 DG counterweight y′ = −gy to reduce p 0 to py 0 ∧ y > 0. 2 DG counter-counterweight z′ = g

2z to reduce y > 0 to yz2 = 1.

3 yz2 = 1 and (after DC with y > 0) py 0 are differential invariants

by construction as (py)′ = p′y − gyp ≥ 0 from premise since y > 0.

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-26
SLIDE 26

ODE Axiomatization: Derived Darboux Rules

Darboux inequalities are DG Q ⊢ p′ ≥ gp p 0 ⊢ [x′ = f (x) & Q]p 0 (g ∈ R[x])

p'=gp t

1

  • v

(1−u2−v2)′ ≥ − 1

2(u2+v2)(1−u2−v2)

. . . ⊢

  • u′ = −v + u

4(1−u2−v2)

v′ = u + v

4(1−u2−v2)

  • 1−u2−v2 > 0

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-27
SLIDE 27

ODE Axiomatization: Derived Darboux Rules

Darboux inequalities are DG Q ⊢ p′ ≥ gp p 0 ⊢ [x′ = f (x) & Q]p 0 (g ∈ R[x])

p'=gp yp=1 y'=-gy t

1

  • v

(1−u2−v2)′ ≥ − 1

2(u2+v2)(1−u2−v2)

. . . ⊢

  • u′ = −v + u

4(1−u2−v2)

v′ = u + v

4(1−u2−v2)

y′ = 1

2(u2+v2)y

  • 1−u2−v2 > 0

(1−u2−v2)y > 0

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-28
SLIDE 28

ODE Axiomatization: Derived Darboux Rules

Darboux inequalities are DG Q ⊢ p′ ≥ gp p 0 ⊢ [x′ = f (x) & Q]p 0 (g ∈ R[x])

p'=gp yp=1 y'=-gy t

1

  • v

(1−u2−v2)′ ≥ − 1

2(u2+v2)(1−u2−v2)

. . . ⊢

  • u′ = −v + u

4(1−u2−v2)

v′ = u + v

4(1−u2−v2)

y′ = 1

2(u2+v2)y

z′ = − 1

4(u2+v2)z

  • 1−u2−v2 > 0

(1−u2−v2)y > 0 yz2 = 1

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 8 / 18

slide-29
SLIDE 29

ODE Axiomatization: Derived Darboux Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 (G ∈ R[x]n×n) Definable p′ for component-wise Lie-derivative w.r.t. ODE

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 9 / 18

slide-30
SLIDE 30

ODE Axiomatization: Derived Darboux Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 (G ∈ R[x]n×n)

Proof Idea.

1 DG counterweight y′ = −G Ty to change p = 0 to p · y = 0. 2 But: p · y = 0 ⇒ p = 0 even if y = 0. 3 Redo: time-varying independent DG matrix Y ′ = −YG with Y p = 0. 4 Y p = 0 ⇒ p = 0 if det Y = 0. 5 DC det Y = 0 which proves by dbx using Liouville’s identity:

det(Y )′ = − tr (G) det(Y )

6 Continuous change of basis Y −1 balancing out motion of p: constant! 7 Continuous change to new evolving variables is sound by DG. Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 9 / 18

slide-31
SLIDE 31

Time is defined so that motion looks simple ≈Poincar´ e

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 10 / 18

slide-32
SLIDE 32

ODE Axiomatization: Derived Invariant Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 11 / 18

slide-33
SLIDE 33

ODE Axiomatization: Derived Invariant Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 Differential Radical Invariants are vdbx Γ, Q ⊢ N−1

i=0 p(i) = 0 Q ⊢ p(N) = N−1 i=0 gip(i)

Γ ⊢ [x′ = f (x) & Q]p = 0

p = 0 p' = 0 p'' = 0 p''' = 0

p = 0 p' = 0 p'' = 0 p''' = 0

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 11 / 18

slide-34
SLIDE 34

ODE Axiomatization: Derived Invariant Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 Differential Radical Invariants are vdbx Γ, Q ⊢ N−1

i=0 p(i) = 0 Q ⊢ p(N) = N−1 i=0 gip(i)

Γ ⊢ [x′ = f (x) & Q]p = 0

p = 0 p' = 0 p'' = 0 p''' = 0

Proof Idea.

by vdbx with G =         1 . . . ... ... . . . . . . . . . ... ... . . . 1 g0 g1 . . . gN−2 gN−1         , p =        p p(1) p(2) . . . p(N−1)       

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 11 / 18

slide-35
SLIDE 35

ODE Axiomatization: Derived Invariant Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 Differential Radical Invariants are vdbx Γ, Q ⊢ N−1

i=0 p(i) = 0 Q ⊢ p(N) = N−1 i=0 gip(i)

Γ ⊢ [x′ = f (x) & Q]p = 0

p = 0 p' = 0 p'' = 0 p''' = 0

p′∗ = 0 N exists

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 11 / 18

slide-36
SLIDE 36

ODE Axiomatization: Derived Invariant Rules

Vectorial Darboux are VDG Q ⊢ p′ = Gp p = 0 ⊢ [x′ = f (x) & Q]p = 0 Differential Radical Invariants are vdbx Γ, Q ⊢ N−1

i=0 p(i) = 0 Q ⊢ p(N) = N−1 i=0 gip(i)

Γ ⊢ [x′ = f (x) & Q]p = 0

p = 0 p' = 0 p'' = 0 p''' = 0

Semialgebraic Invariants are derived p=0 ⊢ p′≥0 .. p=0∧..∧p(N−2)=0 ⊢ p(N−1)≥0 p ≥ 0 ⊢ [x′ = f (x)]p ≥ 0 p′∗ = 0 N exists p′∗ ≥ 0

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 11 / 18

slide-37
SLIDE 37

ODE Axiomatization from Higher Derivatives and Ghosts

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 12 / 18

slide-38
SLIDE 38

ODE Axiomatization: Derived Semialgebraic Rules

Semialgebraic invariants are derived P ⊢

M

  • i=0

m(i)

  • j=0

pij ′∗≥0∧

n(i)

  • j=0

qij ′∗>0

  • ¬P ⊢

N

  • i=0

a(i)

  • j=0

rij ′∗−≥0∧

b(i)

  • j=0

sij ′∗−>0

  • P ⊢ [x′ = f (x)]P

P ≡

M

  • i=0

m(i)

  • j=0

pij ≥ 0 ∧

n(i)

  • j=0

qij > 0

  • ¬P ≡

N

  • i=0

a(i)

  • j=0

rij ≥ 0 ∧

b(i)

  • j=0

sij > 0

  • p′∗=0 ≡

N−1

  • i=0

p(i) = 0 p′∗≥0 ≡ p′∗>0 ∨ p′∗=0 p(N) =

N−1

  • i=0

gip(i) q′∗>0 ≡ q ≥ 0 ∧

  • q = 0 → q′ ≥ 0
  • q = 0 ∧ q′ = 0 → q(2) ≥ 0
  • ∧ . . .

  • q = 0 ∧ q′ = 0 ∧ · · · ∧ q(N−2) = 0 → q(N−1) > 0
  • Definable p′∗− for all/most significant Lie derivatives w.r.t. backwards ODE

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 13 / 18

slide-39
SLIDE 39

ODE Axiomatization: Derived Semialgebraic Rules

Semialgebraic invariants are derived P ⊢

M

  • i=0

m(i)

  • j=0

pij ′∗≥0∧

n(i)

  • j=0

qij ′∗>0

  • ¬P ⊢

N

  • i=0

a(i)

  • j=0

rij ′∗−≥0∧

b(i)

  • j=0

sij ′∗−>0

  • P ⊢ [x′ = f (x)]P

P ≡

M

  • i=0

m(i)

  • j=0

pij ≥ 0 ∧

n(i)

  • j=0

qij > 0

  • ¬P ≡

N

  • i=0

a(i)

  • j=0

rij ≥ 0 ∧

b(i)

  • j=0

sij > 0

  • p′∗=0 ≡

N−1

  • i=0

p(i) = 0 p′∗≥0 ≡ p′∗>0 ∨ p′∗=0 p(N) =

N−1

  • i=0

gip(i) q′∗>0 ≡ q ≥ 0 ∧

  • q = 0 → q′ ≥ 0
  • q = 0 ∧ q′ = 0 → q(2) ≥ 0
  • ∧ . . .

  • q = 0 ∧ q′ = 0 ∧ · · · ∧ q(N−2) = 0 → q(N−1) > 0
  • Seriously?

Fortunately, it’s just a derived rule! Definable p′∗− for all/most significant Lie derivatives w.r.t. backwards ODE

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 13 / 18

slide-40
SLIDE 40

Extended Axiomatization for Semialgebraics

Real Induction [x′ = f (x)]P ↔ ∀y [x′ = f (x) & P ∨ x=y]

  • x=y → P ∧ x′ = f (x) & Px=y
  • x

P r x′ = f(x)

Continuous Existence p > 0 →

  • x′ = f (x) & p > 0
  • x

p > 0 r x′ = f(x) & p > 0

Unique Solutions x′ = f (x) & Q1P1 ∧ x′ = f (x) & Q2P2 → x′ = f (x) & Q1 ∧ Q2(P1 ∨ P2)

x Q1 P r x′ = f(x) & Qi P2 P1 Q1 Q2

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 14 / 18

slide-41
SLIDE 41

ODE Axiomatization: Derived Local Progress Rules

Local Progress Step p > 0 ∨ p = 0 ∧

  • x′ = f (x) & p′ ≥ 0
  • x′ = f (x) & p ≥ 0
  • Local Progress ≥

p′∗ ≥ 0 →

  • x′ = f (x) & p ≥ 0
  • Local Progress >

p′∗ > 0 →

  • x′ = f (x) & p > 0
  • Andr´

e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 15 / 18

slide-42
SLIDE 42

Differential Equation Axiomatization

Theorem (Algebraic Completeness) (LICS’18)

dL calculus is a sound & complete axiomatization of algebraic invariants of polynomial differential equations. They are decidable with a derived axiom (on open Q for completeness): (DRI) [x′ = f (x) & Q]p = 0 ↔

  • Q → p′∗ = 0
  • Theorem (Semialgebraic Completeness)

(LICS’18)

dL calculus with RI is a sound & complete axiomatization of semialgebraic invariants of differential equations. They are decidable with derived axiom (SAI) ∀x (P → [x′ = f (x)]P) ↔ ∀x

  • P → P′∗

∧ ∀x

  • ¬P → (¬P)′∗−

Definable p′∗ is short for all/most significant Lie derivatives w.r.t. ODE Definable p′∗− is w.r.t. backwards ODE. Also for DNF P.

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 16 / 18

slide-43
SLIDE 43

Outline

1

Differential Dynamic Logic

2

Proofs for Differential Equations Differential Invariants / Cuts / Ghosts

3

Completeness for Differential Equation Invariants Darboux are Differential Ghosts Derived Semialgebraic Invariants Real Induction Derived Local Progress Completeness for Invariants

4

Summary

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 16 / 18

slide-44
SLIDE 44

Logical Foundation for Differential Equation Invariants

differential dynamic logic

dL = DL + HP

[α]ϕ ϕ α

1 Poincar´

e: qualitative ODE

2 Complete axiomatization 3 Algebraic ODE invariants 4 Semialgebraic ODE invariants 5 Algebraic hybrid systems 6 Local ODE progress 7 Decide by dL proof/disproof 8 Uniform substitution axioms

Properties

1 MVT 2 Prefix 3 Picard-Lind 4 R-complete 5 Existence 6 Uniqueness 1 Differential invariants 2 Differential cuts 3 Differential ghosts 4 Real induction 5 Continuous existence 6 Unique solutions

Impressive power of differential ghosts

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 17 / 18

slide-45
SLIDE 45

Differential Equation Axiomatization vs. Derived Rules

x w u r x′ = f(x) & Q P w Q x Q w u r x′ = f(x) & Q C w Q x Q w u r x′ = f(x) & Q x P r x′ = f(x) x p > 0 r x′ = f(x) & p > 0 x Q1 P r x′ = f(x) & Qi P2 P1 Q1 Q2

p = 0 p' = 0 p'' = 0 p''' = 0

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 18 / 18

slide-46
SLIDE 46
  • A. Platzer. Logical Foundations of Cyber-Physical Systems. Springer 2018

I Part: Elementary Cyber-Physical Systems

  • 1. Differential Equations & Domains
  • 2. Choice & Control
  • 3. Safety & Contracts
  • 4. Dynamical Systems & Dynamic Axioms
  • 5. Truth & Proof
  • 6. Control Loops & Invariants
  • 7. Events & Responses
  • 8. Reactions & Delays

II Part: Differential Equations Analysis

  • 9. Differential Equations & Differential Invariants
  • 10. Differential Equations & Proofs
  • 11. Ghosts & Differential Ghosts
  • 12. Differential Invariants & Proof Theory

III Part: Adversarial Cyber-Physical Systems 13-16. Hybrid Systems & Hybrid Games IV Part: Comprehensive CPS Correctness

Logical Foundations of Cyber-Physical Systems

André Platzer

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 19 / 18

slide-47
SLIDE 47

Andr´ e Platzer and Yong Kiam Tan. Differential equation axiomatization: The impressive power of differential ghosts. In Anuj Dawar and Erich Gr¨ adel, editors, LICS, pages 819–828, New York, 2018. ACM. doi:10.1145/3209108.3209147. Andr´ e Platzer. Differential-algebraic dynamic logic for differential-algebraic programs.

  • J. Log. Comput., 20(1):309–352, 2010.

doi:10.1093/logcom/exn070. Andr´ e Platzer. The structure of differential invariants and differential cut elimination.

  • Log. Meth. Comput. Sci., 8(4:16):1–38, 2012.

doi:10.2168/LMCS-8(4:16)2012. Andr´ e Platzer. A complete uniform substitution calculus for differential dynamic logic.

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 19 / 18

slide-48
SLIDE 48
  • J. Autom. Reas., 59(2):219–265, 2017.

doi:10.1007/s10817-016-9385-1. Andr´ e Platzer. Logical Foundations of Cyber-Physical Systems. Springer, Switzerland, 2018. URL: http://www.springer.com/978-3-319-63587-3, doi:10.1007/978-3-319-63588-0.

Andr´ e Platzer, Yong Kiam Tan (CMU) Differential Equation Axiomatization LICS’18 19 / 18