TCTL model checking lower/upper-bound Introduction parametric timed - - PowerPoint PPT Presentation

tctl model checking lower upper bound
SMART_READER_LITE
LIVE PREVIEW

TCTL model checking lower/upper-bound Introduction parametric timed - - PowerPoint PPT Presentation

TCTL model checking lower/upper-bound parametric timed automata without invariants TCTL model checking lower/upper-bound Introduction parametric timed automata without Parametric timed automata invariants Model checking with unknown


slide-1
SLIDE 1

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

TCTL model checking lower/upper-bound parametric timed automata without invariants

FORMATS 2018

´ Etienne Andr´ e ∗, Didier Lime ∗∗& Mathias Ramparison∗

*LIPN, Universit´ e Paris 13 **LS2N, ´ Ecole Centrale de Nantes

September 5th, 2018

1/18

slide-2
SLIDE 2

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Outline

Introduction Parametric timed automata Model checking with unknown constants Decision problems for parametric timed automata Contributions U-PTA L/U-PTA Conclusion References

2/18

slide-3
SLIDE 3

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Introduction

◮ Discovering a bug during a test of a system can be very

expensive

◮ Can have dramatical consequences in critical

embedded system: autonomous car, in aeronautics...

3/18

slide-4
SLIDE 4

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Introduction

◮ Discovering a bug during a test of a system can be very

expensive

◮ Can have dramatical consequences in critical

embedded system: autonomous car, in aeronautics...

◮ Need for formal verification to ensure ahead the good

behavior of a system

3/18

slide-5
SLIDE 5

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Model checking

◮ Model of a system:

l1 l2 l3

◮ A property of the system:

l3 is reachable

◮ Check whether the system satisfies the property ◮ Timed Automata [AD94] is a powerful formalism when

all timing constants are known

4/18

slide-6
SLIDE 6

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Model checking with unknown constants

◮ What if all constants are not specified ahead? ◮ Model of a system with parameters:

l1 l2 l3 p1 clock p2 = clock

◮ A property of the system:

l3 is reachable

◮ Compute the values of p1, p2 such that the system

satisfies the property

5/18

slide-7
SLIDE 7

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Example of parametric timed automaton

A parametric timed automaton [AHV93] which models a parametric coffee machine

l1 l2 l3 press: x := 0 y := 0 prepare: y = p1 press again: y 5, x > 1 x := 0 serve: y = p2

◮ Locations : {l1, l2, l3}, clocks : {x, y}, action : {press, press

again, prepare, serve}

◮ Guard(press again) = {y 5 ∧ x 0},

Guard(prepare) = {y = p1}, Guard(serve) = {y = p2}

◮ Reset(press) = {x, y := 0}, Reset(press again) = {x := 0}

6/18

slide-8
SLIDE 8

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Example of parametric timed automaton

A parametric timed automaton [AHV93] which models a parametric coffee machine

l1 l2 l3 press: x := 0 y := 0 prepare: y = p1 press again: y 5, x > 1 x := 0 serve: y = p2

◮ A possible run if p1 = 2, p2 = 3:

  • l1, (0, 0)

press − →

2

  • l2, (0, 0)

press again − →

1.1

  • l2, (0, 1.1)

prepare − →

.9

  • l3, (.9, 2)

serve − →

1

  • l1, (1.9, 3)
  • ◮ The same run is impossible if p1 = 5, p2 = 2.

6/18

slide-9
SLIDE 9

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Flat (no nesting) TCTL decision problems for PTAs

◮ EF-emptiness: is the set of parameter valuations s.t.

there exists a run reaching l in the instantiated TA empty ?

◮ EF-universality: are all parameter valuations s.t. there

exists a run reaching l in the instantiated TA

◮ EG-emptiness: is the set of valuations for which one

infinite or finite maximal runs always remains in a given set of locations empty?

◮ AF-emptiness: is the set of valuations for which all runs

eventually reach a given location empty? (equivalent to EG-universality)

◮ AG-emptiness: is the set of valuations for which all

infinite or finite maximal run always remain in a given set of locations empty?

7/18

slide-10
SLIDE 10

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Challenges for parametric timed automata

◮ EF-emptiness problem: proved undecidable in general

case [AHV93], unbounded integer-valued parameters, (un)bounded rational valued parameters and even with

  • nly one bounded parameter [Mil00]

◮ To recover decidability, we need to add restrictions on

parameters, or restrain the PTA syntax

8/18

slide-11
SLIDE 11

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

L/U-PTA

Lower/upper bound PTAs (L/U-PTAs) introduced in [HRSV02]. Here is an L/U-PTA without invariant.

l1 l2 l3 press: x := 0 y := 0 prepare: y = 5 press again: y p1, x > p2 x := 0 serve: y = 8

Comparison with:

Upper-bound parameter p1.

Lower-bound parameter p2. 9/18

slide-12
SLIDE 12

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

U-PTA

U-PTAs [BL09]: no undecidability result, and almost all decidability results are from L/U-PTAs

◮ Decidability of EF-emptiness and universality for

integer-valued U-PTAs [BL09]

◮ Decidability language preservation synthesis for one

parameter and a deterministic automaton [AM15] Here is a U-PTA without invariant.

l1 l2 l3 press: x := 0 y := 0 prepare: y = 5 press again: y p, x > 1 x := 0 serve: y = 8

Upper-bound parametric guard: y p. 10/18

slide-13
SLIDE 13

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Current results and contributions

Class U-PTAs integer-valued L/U-PTAs without invariant L/U-PTAs PTAs EF [HRSV02] [HRSV02] [HRSV02] [AHV93, Mil00] AF

  • pen
  • pen

[JLR15] [JLR15] EG

  • pen
  • pen

[AL17] [AL17] AG [HRSV02] [HRSV02] [HRSV02] [ALR16a] flat TCTL

  • pen
  • pen

[JLR15] [AHV93] TCTL

  • pen
  • pen

[JLR15] [AHV93]

Table: Decidability of the emptiness problems for PTAs and subclasses Contributions:

◮ Undecidability of non-flat TCTL (with nesting) for unbounded

U-PTA without invariant

◮ Undecidability of non-flat TCTL for bounded U-PTAs without

invariant

◮ Decidability of EG-emptiness/universality (in PSPACE) for

integer-valued L/U-PTAs without invariant

11/18

slide-14
SLIDE 14

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

U-PTA

U-PTAs without invariant with rational-valued parameters

  • ver dense time.

12/18

slide-15
SLIDE 15

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

U-PTA

U-PTAs without invariant with rational-valued parameters

  • ver dense time.

Theorem (1)

Non-flat-TCTL is undecidable for rational-valued U-PTAs without invariant. Proof sketch: we prove that the EGAF=0-emptiness problem is undecidable for rational-valued U-PTAs without invariant, using a reduction from the halting problem of a two counter machine

l l ′ l ′′ locerror x1 a, t b y := 0 y = 0, t b x1 := 0 x1 a, y > 0 12/18

slide-16
SLIDE 16

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

U-PTA

U-PTAs without invariant with bounded rational-valued parameters over dense time Motivation:

◮ it is impossible to simulate a bounded U-PTA using a

U-PTA [ALR16b],

◮ and EG-emptiness is decidable for bounded L/U-PTAs,

but undecidable for L/U-PTAs [AL17].

13/18

slide-17
SLIDE 17

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

U-PTA

U-PTAs without invariant with bounded rational-valued parameters over dense time Motivation:

◮ it is impossible to simulate a bounded U-PTA using a

U-PTA [ALR16b],

◮ and EG-emptiness is decidable for bounded L/U-PTAs,

but undecidable for L/U-PTAs [AL17].

Theorem (2)

Non-flat-TCTL is undecidable for bounded rational-valued U-PTAs without invariant. Proof sketch: we prove that the EGAF=0-emptiness problem is undecidable for bounded rational-valued U-PTAs without invariant, using a reduction from the boundedness problem

  • f a two counter machine

13/18

slide-18
SLIDE 18

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Current results

Class U-PTAs integer-valued L/U-PTAs without invariant L/U-PTAs PTAs EF [HRSV02] [HRSV02] [HRSV02] [AHV93, Mil00] AF

  • pen
  • pen

[JLR15] [JLR15] EG

  • pen
  • pen

[AL17] [AL17] AG [HRSV02] [HRSV02] [HRSV02] [ALR16a] flat TCTL

  • pen
  • pen

[JLR15] [AHV93] TCTL Theorem 1 Theorem 1 [JLR15] [AHV93]

Table: Decidability of the emptiness problems for PTAs and subclasses

14/18

slide-19
SLIDE 19

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

L/U-PTA

L/U-PTAs without invariant with integer-valued parameters

  • ver dense time.

15/18

slide-20
SLIDE 20

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

L/U-PTA

L/U-PTAs without invariant with integer-valued parameters

  • ver dense time.

Theorem (3)

The EG-emptiness/universality problems are PSPACE-complete for integer-valued L/U-PTAs without invariant.

Corollary

Flat-TCTL is decidable for integer-valued L/U-PTAs without invariant (using [HRSV02]).

15/18

slide-21
SLIDE 21

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

L/U-PTA

Proof sketch: We reduce this problem to reachability of a location Is there possibly a deadlock ?

l l ′ x1 a 16/18

slide-22
SLIDE 22

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

L/U-PTA

Proof sketch: We reduce this problem to reachability of a location Remove transition and add self loop

l l ′

And then check whether there is an infinite run

16/18

slide-23
SLIDE 23

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Summary of contributions and conclusion

Class U-PTAs integer-valued L/U-PTAs without invariant L/U-PTAs PTAs EF [HRSV02] [HRSV02] [HRSV02] [AHV93, Mil00] AF

  • pen

Theorem 3 [JLR15] [JLR15] EG

  • pen

Theorem 3 [AL17] [AL17] AG [HRSV02] [HRSV02] [HRSV02] [ALR16a] flat TCTL

  • pen

Theorem 3 [JLR15] [AHV93] TCTL Theorem 1 Theorem 1 [JLR15] [AHV93]

Table: Decidability of the emptiness problems for PTAs and subclasses

◮ Non-flat-TCTL is undecidable for U-PTAs without invariant

(bounded or not).

◮ EG-emptiness and universality (first non trivial subclass of

PTAs) is decidable for integer-valued L/U-PTAs without invariant.

17/18

slide-24
SLIDE 24

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Summary of contributions and conclusion

Class U-PTAs integer-valued L/U-PTAs without invariant L/U-PTAs PTAs EF [HRSV02] [HRSV02] [HRSV02] [AHV93, Mil00] AF

  • pen

Theorem 3 [JLR15] [JLR15] EG

  • pen

Theorem 3 [AL17] [AL17] AG [HRSV02] [HRSV02] [HRSV02] [ALR16a] flat TCTL

  • pen

Theorem 3 [JLR15] [AHV93] TCTL Theorem 1 Theorem 1 [JLR15] [AHV93]

Table: Decidability of the emptiness problems for PTAs and subclasses Future work:

◮ Where exactly the undecidability starts (in particular whether

EG and AF are decidable for U-PTAs with invariants or real-valued parameters), which remains open,

◮ whether our proofs for bounded U-PTAs can be extended

  • ver bounded time,

◮ whether the same results hold for L-PTAs (lower-bound

PTAs).

17/18

slide-25
SLIDE 25

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Advertisement: Paris summer school 2019

Paris Summer School 2019 in Specification and Verification of Critical Systems 1st to 19th July 2019

◮ For Chinese Master students ◮ Supported by Campus France / French Embassy in

Beijing

◮ Scientific organization: most Paris labs in verification

(Sorbonne Universit´ e, ENS Paris-Saclay, Le CNAM, Universit´ e Paris-Est Cr´ eteil Val de Marne, Sorbonne Paris Cit´

  • e. . . )

More info:

www.lipn13.fr/paris-2019/

18/18

slide-26
SLIDE 26

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

References

Rajeev Alur and David L. Dill. A theory of timed automata. Theoretical Computer Science, 126(2):183–235, April 1994. Rajeev Alur, Thomas A. Henzinger, and Moshe Y. Vardi. Parametric real-time reasoning. In S. Rao Kosaraju, David S. Johnson, and Alok Aggarwal, editors, STOC, pages 592–601, New York, NY, USA, 1993. ACM. ´ Etienne Andr´ e and Didier Lime. Liveness in L/U-parametric timed automata. In Alex Legay and Klaus Schneider, editors, ACSD, pages 9–18. IEEE, 2017. ´ Etienne Andr´ e, Didier Lime, and Olivier H. Roux. Decision problems for parametric timed automata. In Kazuhiro Ogata, Mark Lawford, and Shaoying Liu, editors, ICFEM, volume 10009 of Lecture Notes in Computer Science, pages 400–416. Springer, 2016. ´ Etienne Andr´ e, Didier Lime, and Olivier H. Roux. On the expressiveness of parametric timed automata. In Martin Fr¨ anzle and Nicolas Markey, editors, FORMATS, volume 9984, pages 19–34. Springer, 2016. ´ Etienne Andr´ e and Nicolas Markey. Language preservation problems in parametric timed automata. In FORMATS, volume 9268, pages 27–43. Springer, 2015. Laura Bozzelli and Salvatore La Torre. Decision problems for lower/upper bound parametric timed automata. Formal Methods in System Design, 35(2):121–151, 2009. Thomas Hune, Judi Romijn, Mari¨ elle Stoelinga, and Frits W. Vaandrager. Linear parametric model checking of timed automata. Journal of Logic and Algebraic Programming, 52-53:183–220, 2002. Aleksandra Jovanovi´ c, Didier Lime, and Olivier H. Roux. 19/18

slide-27
SLIDE 27

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Clock regions

x y R1 R2 R4 R3 R6 R5

◮ The corner point: R1 = {(4, 4)} ◮ The vertical line: R2 = {(x, y) | x = 2, 0 < y < 1} ◮ The horizontal line: R3 = {(x, y) | y = 3, 1 < x < 2} ◮ The diagonal: R4 = {(x, y) | x = y − 3, 4 < y < 5} ◮ The upward triangle: R5 = {(x, y) | 0 < x < y − 1, 1 < y < 2} ◮ The downward triangle: R6 = {(x, y) | y + 1 < x < 4, 2 < y < 3}

20/18

slide-28
SLIDE 28

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Clock regions

Two clocks x, y, max constants cx = 2, cy = 1. Time successors of the blue region

{0 < y < 1, 0 < y < x − 1} different of itself: four regions in

green: {0 < y < 1, x = 2}, {0 < y < 1, x > 2}, {y = 1, x > 2} and {y > 1, x > 2} y 1 1 2 x

21/18

slide-29
SLIDE 29

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Using regions for parametric timed automata ?

l0 l1 l2 x = p → x := 0 y = 1 ∧ x = p

In l1: (x, y) = (0, p) But after letting some time elapse, depending on the value of 0 < p < 1 we reach different regions:

◮ region y = 1, 0 < x < p if 1 > p > 1 2

y 1 1 x p p

22/18

slide-30
SLIDE 30

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Using regions for parametric timed automata ?

l0 l1 l2 x = p → x := 0 y = 1 ∧ x = p

In l1: (x, y) = (0, p) But after letting some time elapse, depending on the value of 0 < p < 1 we access different regions:

◮ region y = 1, x = p if p = 1 2

y 1 1 x p p

23/18

slide-31
SLIDE 31

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Using regions for parametric timed automata ?

l0 l1 l2 x = p → x := 0 y = 1 ∧ x = p

In l1: (x, y) = (0, p) But after letting some time elapse, depending on the value of 0 < p < 1 we access different regions:

◮ region p < y < 1, x = p if p < 1 2

y 1 1 x p p

24/18

slide-32
SLIDE 32

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Proof sketch U-PTAs

li li

1

li

2

li

3

li

4

locerror li

5

li

6

li

7

lj li

8

li

9

li

10

li

11

z = 1, t b z := 0 x1 a, t b y := 0 y = 0, t b x1 := 0 x1 a, y > 0 x2 a, t b y := 0 x2 a, y > 0 y = 0, t b x2 := 0 x2 = 1, t b x2 := 0 z a, t b y := 0 z a, y > 0 y = 0, t b z := 0 x2 a, t b y := 0 y = 0, t b x2 := 0 x2 a, y > 0 x2 = 1, t b x2 := 0 x1 a, t b y := 0 x

1

  • a

, y > y = 0, t b x1 := 0

25/18

slide-33
SLIDE 33

TCTL model checking lower/upper-bound parametric timed automata without invariants Introduction Parametric timed automata

Model checking with unknown constants Decision problems for parametric timed automata

Contributions

U-PTA L/U-PTA

Conclusion References

Proof sketch bounded U-PTAs

li li

1

li

2

li

3

locerror li

4

lj li

5

li

6

z = 0 x

2

= 1 x

2

: = x1 a + 1 y := 0 x

1

  • a

+ 1 , y > y = x

1

: = z = 1 z := 0 x

1

  • a

+ 1 y : = y = 0 x1 := 0 x

1

  • a

+ 1 , y > x

2

= 1 x

2

: =

26/18