Validation of Automotive Control Applications using Formal Methods - - PowerPoint PPT Presentation

▶
validation of automotive control applications using
SMART_READER_LITE
LIVE PREVIEW

Validation of Automotive Control Applications using Formal Methods - - PowerPoint PPT Presentation

Dec 25, 2023 786 likes •1.44k views

Validation of Automotive Control Applications using Formal Methods and metamodeling techniques Simone Silvetti, Esteco Spa & University Udine Mariapia Marchi, Esteco Spa www.caeconference.com MDB ( M odel B ased D evelopment)

slide-1
SLIDE 1

Validation of Automotive Control Applications using Formal Methods and metamodeling techniques

❖ Simone Silvetti, Esteco Spa & University Udine ❖ Mariapia Marchi, Esteco Spa

slide-2
SLIDE 2

www.caeconference.com

MDB (Model Based Development)

❖ process aimed at designing complex systems ❖ cost reduction ❖ reduce development time

2
slide-3
SLIDE 3

www.caeconference.com

MDB (Model Based Development)

3 17 - 18 October 2016 International CAE Conference
slide-4
SLIDE 4

www.caeconference.com

MDB (Model Based Development)

4 17 - 18 October 2016 International CAE Conference
slide-5
SLIDE 5

www.caeconference.com

MDB (Model Based Development)

5 17 - 18 October 2016 International CAE Conference
slide-6
SLIDE 6

www.caeconference.com

MDB (Model Based Development)

6 17 - 18 October 2016 International CAE Conference
slide-7
SLIDE 7

www.caeconference.com

MDB (Model Based Development)

7 17 - 18 October 2016 International CAE Conference
slide-8
SLIDE 8

www.caeconference.com

MDB (Model Based Development)

8 17 - 18 October 2016 International CAE Conference
slide-9
SLIDE 9

www.caeconference.com

Validation Process

9 17 - 18 October 2016 International CAE Conference
slide-10
SLIDE 10

www.caeconference.com

Validation Process

10 10

❖ Use of block diagram tools (Simulink, Gt suite) ❖ Powerful Tools but complex

17 - 18 October 2016 International CAE Conference
slide-11
SLIDE 11

www.caeconference.com

Validation Process

11 11

❖ Use of block diagram tools (Simulink, Gt suite) ❖ Use of natural languages ❖ Involves time events... ❖ Powerful Tools but complex ❖ Not rigorous ❖ Not Machine interpretable

17 - 18 October 2016 International CAE Conference
slide-12
SLIDE 12

www.caeconference.com

Validation Process

12 12

❖ Use of block diagram tools (Simulink, Gt suite) ❖ Use of natural languages ❖ Involves time events... ❖ Powerful Tools but complex ❖ Not rigurous ❖ Not Machine interpretable

17 - 18 October 2016 International CAE Conference
slide-13
SLIDE 13

www.caeconference.com

Validation Process

13 13

❖ Use of block diagram tools (Simulink, Gt suite) ❖ Use of natural languages ❖ Involves time events... ❖ Powerful Tools but complex ❖ Not rigurous ❖ Not Machine interpretable

FORMAL METHODS !

✓

17 - 18 October 2016 International CAE Conference
slide-14
SLIDE 14

www.caeconference.com

Validation Process

14 14 17 - 18 October 2016 International CAE Conference
slide-15
SLIDE 15

www.caeconference.com

Validation Process

15 15 17 - 18 October 2016 International CAE Conference
slide-16
SLIDE 16

www.caeconference.com

Validation Process

16 16

φ

17 - 18 October 2016 International CAE Conference
slide-17
SLIDE 17

www.caeconference.com

Validation Process

17 17

φ

17 - 18 October 2016 International CAE Conference
slide-18
SLIDE 18

www.caeconference.com

Validation Process

18 18

“If the engine speed (w) is always less than k1 then vehicle speed (v) can not exceed k2 in less than T sec” ᅟᅠᆨ(F[0,T] (v ≥ k2) ⋀ G(w ≤ k1))

φ

17 - 18 October 2016 International CAE Conference
slide-19
SLIDE 19

www.caeconference.com

Robustness Semantics

19 19

φ ⊧ ?

17 - 18 October 2016 International CAE Conference
slide-20
SLIDE 20

www.caeconference.com

Robustness Semantics

20 20

φ ⊧ ?

Boolean yes/no

17 - 18 October 2016 International CAE Conference

k F( f>k )

slide-21
SLIDE 21

www.caeconference.com

Robustness Semantics

21 21

φ ⊧ ?

Boolean Robustness yes/no +30 / -30

More Information!

17 - 18 October 2016 International CAE Conference

+30 k F( f>k )

slide-22
SLIDE 22

www.caeconference.com

The goal

22 22

f M M(f)

17 - 18 October 2016 International CAE Conference
slide-23
SLIDE 23

www.caeconference.com

The goal

23 23

f M M(f) min [M(f), φ ]

f ∈ F

The optimization Problem

R =

17 - 18 October 2016 International CAE Conference
slide-24
SLIDE 24

www.caeconference.com

The goal

24 24

f M M(f) min [M(f), φ ]

f ∈ F

The optimization Problem

R =

R

Counterexample Safe!

≤ 0 ≥ 0

17 - 18 October 2016 International CAE Conference
slide-25
SLIDE 25

www.caeconference.com

The optimization process

slide-26
SLIDE 26

www.caeconference.com

26

The optimization process

Challenges

❖

Low number of model execution

❖

Inputs are functions (temporal series)!!

slide-27
SLIDE 27

www.caeconference.com

27

The optimization process

Challenges

❖

Low number of model execution

❖

Inputs are functions (temporal series)!!

slide-28
SLIDE 28

www.caeconference.com

28

The optimization process

Challenges

❖

Low number of model execution

❖

Inputs are functions (temporal series)!! GP-UCB Adaptive Control Point Parametrization

slide-29
SLIDE 29

www.caeconference.com

The Control Point Parametrization

29

Fix the times interpolation

slide-30
SLIDE 30

www.caeconference.com

The Control Point Parametrization

30

Fix the times interpolation n Control Points n Variable to optimize

slide-31
SLIDE 31

www.caeconference.com

The Control Point Parametrization

31

Fix the times interpolation n Control Points n Variable to optimize

slide-32
SLIDE 32

www.caeconference.com

The adaptive Control Point Param.

32

n Control Points 2n Variable to optimize interpolation

slide-33
SLIDE 33

www.caeconference.com

33

Doubled the variables

Problem

Increase the expressivity but...

slide-34
SLIDE 34

www.caeconference.com

34

Doubled the variables

Problem Solution

GP-UCB Optimizer

Increase the expressivity but...

slide-35
SLIDE 35

www.caeconference.com

GP-UCB

35 17 - 18 October 2016 International CAE Conference
slide-36
SLIDE 36

www.caeconference.com

GP-UCB

36 17 - 18 October 2016 International CAE Conference
slide-37
SLIDE 37

www.caeconference.com

GP-UCB

37 17 - 18 October 2016 International CAE Conference
slide-38
SLIDE 38

www.caeconference.com

GP-UCB

38

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-39
SLIDE 39

www.caeconference.com

GP-UCB

39

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-40
SLIDE 40

www.caeconference.com

GP-UCB

40

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-41
SLIDE 41

www.caeconference.com

GP-UCB

41

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-42
SLIDE 42

www.caeconference.com

GP-UCB

42

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-43
SLIDE 43

www.caeconference.com

GP-UCB

43

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-44
SLIDE 44

www.caeconference.com

GP-UCB

44

P(x,y)

17 - 18 October 2016 International CAE Conference
slide-45
SLIDE 45

www.caeconference.com

45

Reduce Input Space Doubled the variables

slide-46
SLIDE 46

www.caeconference.com

Schema

46

GP - UCB N

  • Rob. ?

✓

N++

17 - 18 October 2016 International CAE Conference
slide-47
SLIDE 47

www.caeconference.com

47

Input Space

Adaptive Idea

slide-48
SLIDE 48

www.caeconference.com

48

Input Space

Adaptive Idea

1

slide-49
SLIDE 49

www.caeconference.com

49

Input Space

Adaptive Idea

2

slide-50
SLIDE 50

www.caeconference.com

50

Input Space

Adaptive Idea

2

slide-51
SLIDE 51

www.caeconference.com

51

Input Space

Adaptive Idea

2

slide-52
SLIDE 52

www.caeconference.com

52

Input Space

Adaptive Idea

3

slide-53
SLIDE 53

www.caeconference.com

53

Input Space

Adaptive Idea

3

slide-54
SLIDE 54

www.caeconference.com

54

Input Space

Adaptive Idea

4

slide-55
SLIDE 55

www.caeconference.com

55

Input Space

Adaptive Idea

slide-56
SLIDE 56

www.caeconference.com

56

Automatic transmission

slide-57
SLIDE 57

www.caeconference.com

57

Automatic transmission

slide-58
SLIDE 58

www.caeconference.com

58

Automatic transmission

69 blocks: 2 integrators, 3 look-up tables, 3 2D look-up tables, Stateflow Chart

slide-59
SLIDE 59

www.caeconference.com

59

Results

17 - 18 October 2016 International CAE Conference
slide-60
SLIDE 60

www.caeconference.com

60

aCPP reduces minimum number of evaluations by 50-70%

GP-UCB is slow.

Results

17 - 18 October 2016 International CAE Conference

✓

slide-61
SLIDE 61

www.caeconference.com

61

Results

Time = {#Simulations} x {Simulation Time} + {Optimizer time}

GP-UCB is slow

17 - 18 October 2016 International CAE Conference
slide-62
SLIDE 62

www.caeconference.com

62

Results

Time = {#Simulations} x {Simulation Time} + {Optimizer time}

GP-UCB is slow

Future work

❖ from Matlab to Java (parallelization) ❖ multi-objective approach ❖ using fmi as simulator

17 - 18 October 2016 International CAE Conference
slide-63
SLIDE 63

www.caeconference.com

Acknowledges

Esteco

17 - 18 October 2016 International CAE Conference 63

Alberto Policriti Luca Bortolussi

slide-64
SLIDE 64

www.caeconference.com

64

….and use Formal Methods