Presentation is loading. Please wait.

Presentation is loading. Please wait.

Formal methods & Tools UCb CUPPAAL CUPPAAL Efficient Minimum-Cost Reachability for Linearly Priced Timed Automata Gerd Behrman, Ed Brinksma, Ansgar Fehnker,

Similar presentations


Presentation on theme: "Formal methods & Tools UCb CUPPAAL CUPPAAL Efficient Minimum-Cost Reachability for Linearly Priced Timed Automata Gerd Behrman, Ed Brinksma, Ansgar Fehnker,"— Presentation transcript:

1 Formal methods & Tools UCb CUPPAAL CUPPAAL Efficient Minimum-Cost Reachability for Linearly Priced Timed Automata Gerd Behrman, Ed Brinksma, Ansgar Fehnker, Thomas Hune, Kim Larsen, Paul Pettersson, Judi Romijn, Frits Vaandrager

2 VHS meeting Kim G. Larsen UCb 2 Overview 1.Introduction 2.Linear Priced Timed Automata 3.Priced Zones and Facets 4.Operations on Priced Zones 5.Algorithm 6.First Experimental Findings 7.Conclusion

3 VHS meeting Kim G. Larsen UCb 3 Observation Many scheduling problems can be phrased naturally as reachability problems for timed automata! INTRODUCTION

4 VHS meeting Kim G. Larsen UCb 4 Observation Many scheduling problems can be phrased naturally as reachability problems for timed automata! UNSAFE SAFE At most 2 crossing at a time Need torch At most 2 crossing at a time Need torch Mines Can they make it within 60 minutes ? Can they make it within 60 minutes ? INTRODUCTION

5 VHS meeting Kim G. Larsen UCb 5 Observation Many scheduling problems can be phrased naturally as reachability problems for timed automata! UNSAFE SAFE Mines INTRODUCTION

6 VHS meeting Kim G. Larsen UCb 6 Steel Production Plant Machine 1 Machine 2Machine 3 Machine 4Machine 5 Buffer Continuos Casting Machine Storage Place Crane B Crane A zA. Fehnker, T. Hune, K. G. Larsen, P. Pettersson zCase study of Esprit-LTR project VHS zPhysical plant of SIDMAR located in Gent, Belgium. zPart between blast furnace and hot rolling mill. Objective: model the plant, obtain schedule and control program for plant. Lane 1 Lane 2 INTRODUCTION

7 VHS meeting Kim G. Larsen UCb 7 Batch Processing Plant (VHS) hbrine water store mbrine heat water heater cooling water pump cooling water water salt INTRODUCTION

8 VHS meeting Kim G. Larsen UCb 8 Earlier work zAsarin & Maler (1999) Time optimal control using backwards fixed point computation zVHS consortium (1999) Steel plant and chemical batch plant case studies zNiebert, Tripakis & Yovine (2000) Minimum-time reachability using forward reachability zBehrmann, Fehnker et all (2000) Minimum-time reachability using branch-and-bound INTRODUCTION

9 VHS meeting Kim G. Larsen UCb 9 Advantages Easy and flexible modeling of systems whole range of verification techniques becomes available Controller/Program synthesis Disadvantages Existing scheduling approaches perform somewhat better Our goal See how far we get; Integrate model checking and scheduling theory. INTRODUCTION

10 VHS meeting Kim G. Larsen UCb 10 More general cost function zIn scheduling theory one is not just interested in shortest schedules; also other cost functions are considered zThis leads us to introduce a model of linear priced timed automata which adds prices to locations and transitions zThe price of a transition gives the cost of taking it, and the price of a location specifies the cost per time unit of staying there. INTRODUCTION

11 Formal methods & Tools UCb Linearly Priced Timed Automata

12 VHS meeting Kim G. Larsen UCb 12 Example PRICED AUTOMATA

13 VHS meeting Kim G. Larsen UCb 13 EXAMPLE : Optimal rescue plan for important persons (Presidents and Actors) UNSAFE SAFE Mines GORECLINTON BUSH DIAZ OPTIMAL PLAN HAS ACCUMULATED COST=195 and TOTAL TIME=65! PRICED AUTOMATA

14 VHS meeting Kim G. Larsen UCb 14 Definition PRICED AUTOMATA

15 VHS meeting Kim G. Larsen UCb 15 Definition PRICED AUTOMATA

16 VHS meeting Kim G. Larsen UCb 16 Example of execution PRICED AUTOMATA

17 VHS meeting Kim G. Larsen UCb 17 Cost zThe cost of a finite execution is the sum of the prices of all the transitions occuring in it zThe minimal cost of a location is the infimum of the costs of the finite executions ending in the location zThe minimum-cost problem for LPTAs is the problem to compute the minimal cost of a given location of a given LPTA zIn the example below, mincost(C ) = 7 PRICED AUTOMATA ? DECIDABILITY ?

18 Formal methods & Tools UCb Priced Zones

19 VHS meeting Kim G. Larsen UCb 19 Zones Operations PRICED ZONES

20 VHS meeting Kim G. Larsen UCb 20 Canonical Datastructure for Zones Difference Bounded Matrices x1-x2<=4 x2-x1<=10 x3-x1<=2 x2-x3<=2 x0-x1<=3 x3-x0<=5 x1-x2<=4 x2-x1<=10 x3-x1<=2 x2-x3<=2 x0-x1<=3 x3-x0<=5 x1x2 x3x x1x2 x3x Shortest Path Closure O(n^3) Bellman’58, Dill’89 PRICED ZONES

21 VHS meeting Kim G. Larsen UCb 21 New Canonical Datastructure Minimal collection of constraints x1-x2<=4 x2-x1<=10 x3-x1<=2 x2-x3<=2 x0-x1<=3 x3-x0<=5 x1-x2<=4 x2-x1<=10 x3-x1<=2 x2-x3<=2 x0-x1<=3 x3-x0<=5 x1x2 x3x x1x2 x3x x1x2 x3x Shortest Path Closure O(n^3) Shortest Path Reduction O(n^3) 3 Space worst O(n^2) practice O(n) RTSS 1997 PRICED ZONES

22 VHS meeting Kim G. Larsen UCb 22 Priced Zone PRICED ZONES x y 4 2 Z

23 VHS meeting Kim G. Larsen UCb 23 Reset x y 4 2 Z PRICED ZONES

24 VHS meeting Kim G. Larsen UCb 24 Reset x y 4 2 Z {y}Z PRICED ZONES

25 VHS meeting Kim G. Larsen UCb 25 Reset x y 4 2 Z {y}Z4 PRICED ZONES

26 VHS meeting Kim G. Larsen UCb 26 Reset x y 4 2 Z {y}Z4 1 PRICED ZONES 2 A split of {y}Z 4

27 VHS meeting Kim G. Larsen UCb 27 Facets The solution PRICED ZONES

28 VHS meeting Kim G. Larsen UCb 28 OPERATIONS ON PZONES

29 VHS meeting Kim G. Larsen UCb 29 Delay x y 4 3 Z PRICED ZONES

30 VHS meeting Kim G. Larsen UCb 30 Delay x y 4 3 Z Delay in a location with cost-rate PRICED ZONES

31 VHS meeting Kim G. Larsen UCb 31 Delay x y 4 3 Z PRICED ZONES A split of

32 VHS meeting Kim G. Larsen UCb 32 Facets The solution PRICED ZONES

33 VHS meeting Kim G. Larsen UCb 33 OPERATIONS ON PZONES

34 VHS meeting Kim G. Larsen UCb 34 Optimal Forward Reachability Example PRICED ZONES

35 VHS meeting Kim G. Larsen UCb 35 OPERATIONS ON PZONES

36 VHS meeting Kim G. Larsen UCb 36 OPERATIONS ON PZONES

37 Formal methods & Tools UCb Algorithm

38 VHS meeting Kim G. Larsen UCb 38 Branch & Bound Algorithm ALGORITHM

39 VHS meeting Kim G. Larsen UCb 39 ALGORITHM

40 VHS meeting Kim G. Larsen UCb 40 ALGORITHM

41 Formal methods & Tools UCb Experiments

42 VHS meeting Kim G. Larsen UCb 42 EXAMPLE : Optimal rescue plan for important persons (Presidents and Actors) UNSAFE SAFE Mines GORECLINTON BUSH DIAZ OPTIMAL PLAN HAS ACCUMULATED COST=195 and TOTAL TIME=65! EXPERIMENTS

43 VHS meeting Kim G. Larsen UCb 43 Experiments MC Order COST-rates SCHEDULE COSTTIME #Expl#Pop’d G5G5 C 10 B 20 D 25 Min Time CG> G C CG> G G GD> G G CG> G C CD> C C BD> B C time< EXPERIMENTS

44 VHS meeting Kim G. Larsen UCb 44 Optimal Broadcast Router1 Router2 Router3 Router4 A B Given particular subscriptions, what is the cheapest schedule for broadcasting k? Given particular subscriptions, what is the cheapest schedule for broadcasting k? k=1k=0 costA 1, costB 1 costA 2, costB 2 costA 3, costB 3 costA 4, costB 4 Basecost EXPERIMENTS costB 1 costA 1 3 sec 5 sec

45 VHS meeting Kim G. Larsen UCb 45 Experimental Results COST-rates SCHEDULE COSTTIME #Expl BCR1R1 R2R2 R3R3 R4R4 Min Time 1>3(B) ; ( 3>4(B) | 1>2(A) ) :31:31:31:31:31:31:31:3 1>4(A) ; 3>4(A) ; 4>2(A) >3(B) ; ( 3>4(B) | 1>2(A) ) :30 5 :15 1:31:36:26:2 1>3(A) ; 3>2(A) ; 3>4(A) >4(A) ; 4>3(B) ; 4>2(B) >4(B) ; ( 1>3(A) | 4>2(B) ) t<=10 1>4(B) ; 4>2(B) ; 4>3(B) t<=8 1>4(B) ; ( 1>3(A) | 4>2(B) ) EXPERIMENTS

46 VHS meeting Kim G. Larsen UCb 46 Scaling Up ? z# Schedules x4 routers: 120 x5 routers: x6 routers: ?????????? zFinding Feasible Schedule using UPPAAL (6 routers) x expl. symb. st. (with Active Clock Reduction) zMinimum Time Schedule (6 routers) x using Minimum Time Reachability (Ansgar) x using Minimum Cost Reachability (BC=1, all other cost=0) time optimal schedule takes 12 seconds. EXPERIMENTS

47 VHS meeting Kim G. Larsen UCb 47 Current & Future Work IMPLEMENTATION – thorough analysis  Applications – (Gossing Girls, Production Plant)  Generalization Minimum Cost Reachability under timing constraints avoiding certain states  Minimum Time Reachability under cost constraints  Maximum Cost between two types of states  Relationships to Reward Models Parameterized Extension  Extensions to Optimal Controllability


Download ppt "Formal methods & Tools UCb CUPPAAL CUPPAAL Efficient Minimum-Cost Reachability for Linearly Priced Timed Automata Gerd Behrman, Ed Brinksma, Ansgar Fehnker,"

Similar presentations


Ads by Google