nitin kumar yadav rmit university, melbourne nitin.yadav@student.rmit.au

Nitin Kumar YadavRMIT University, Melbourne

nitin.yadav@student.rmit.edu.au

Minor thesis for semester 2, 2009, under the supervision of Dr. Sebastian Sardina, RMIT university

Implementation and analysis of simulation based techniques for behavior composition

Implementation and analysis of simulation based techniques for behavior composition

Behavior CompositionSimulationTechniquesImplementation Analysis

Contents

Behavior Composition

3

What is a behavior ?

• Behavior – Logic of a machine– Web service– Stand alone component

• Abstracted as finite transition systems• Available behaviors can be non-deterministic

B1 B2

Behavior Composition

4

Combining available behaviors to realize a target behavior

Available behaviors

Target behavior(virtual)

T1

Can we realize T1 by composing B1 and B2 ?

B1 B2

Behavior Composition

5

Combined finite transition system of available behaviors

‘composed’ transition system of available behaviors

B1

B2

Asynchronous product of B1 and B2

Behavior Composition

6

Combined finite transition system of available behaviors

B1

B2

Asynchronous product of B1 and B2

‘composed’ transition system of available behaviors

Behavior Composition

7

Combined finite transition system of available behaviors

B1

B2

Asynchronous product of B1 and B2

Can this behave like the target system ?

Simulation

8

• A transition system T1 simulates another transition system T2 iff T1 can ‘mimic’ all the states of T2

• A state in the available system mimics another state in the target system if:– It can do all the actions that the target state can do– The successor state in the available system as a result of

such an action simulates the resulting state in the target system

• Simulation is a relation of states of the composed system and the states of the target behavior which can be ‘mimicked’.

t

Simulation

9

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S2,S1>, <T2>}…

Simulation Relation

simulation relationis a solution to the behavior

Composition problem !

How to calculate it ?

Techniques

10

Two approaches for behavior composition

• Regression based approach [Sardina,Patrizi & De Giacomo, KR 2008]

• Progression based approach [Stroeder & Pagnucco, 2009, IJCAI 2009]

Proceedings of Principles of Knowledge Representation and Reasoning (KR), pages 640-650, Sydney, Australia, September 2008. AAAI Press.

Accepted for the IJCAI 2009

Techniques

11

Regression based approach [Sardina, Patrizi, De Giacomo]

• Assume each state in the available system simulates each state in the target system

• Iteratively remove non-conformant links which don’t’ follow the simulation definition i.e., – Can not perform the actions which can be requested

in the matching target state– The successor state of the action does not follow the

above rule

• Stop when no more links can be removed

t

Regression based approach

12

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S1,S1>, <T2>}{<S1,S1>, <T3>}{<S2,S1>, <T1>}{<S2,S1>, <T2>}{<S2,S1>, <T3>}{<S2,S2>, <T1>}{<S2,S2>, <T2>}{<S2,S2>, <T3>}{<S1,S2>, <T1>}{<S1,S2>, <T2>}{<S1,S2>, <T3>}

Assume each state from availablebehaviors simulates each stateIn the target system

t

Regression based approach

13

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S1,S1>, <T2>}{<S1,S1>, <T3>}{<S2,S1>, <T1>}{<S2,S1>, <T2>}{<S2,S1>, <T3>}{<S2,S2>, <T1>}{<S2,S2>, <T2>}{<S2,S2>, <T3>}{<S1,S2>, <T1>}{<S1,S2>, <T2>}{<S1,S2>, <T3>}

Each Cycle : step 1 – remove the States which can not perform the Actions of the linked target state

t

Regression based approach

14

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S1,S1>, <T2>}{<S2,S1>, <T1>}{<S2,S1>, <T2>}{<S2,S2>, <T1>}{<S2,S2>, <T2>}{<S2,S2>, <T3>}{<S1,S2>, <T1>}{<S1,S2>, <T3>}

Each Cycle : step 2 – remove the States whose successor states are not in the simulation relation

X

Continue till no more links can beremoved

Techniques

15

Progression based approach [Stroder & Pagnucco]

• Start from the initial state• Iteratively add conformant links between the

states of the composed system and the target system

• Stop when no more links can be added

t

Progression based approach

16

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S2,S1>, <T1>}{<S2,S2>, <T1>}{<S1,S2>, <T1>}

Start from states those ‘canMimic the initial state

t

Progression based approach

17

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S2,S1>, <T1>}{<S2,S2>, <T1>}{<S1,S2>, <T1>}

{<S2,S1>, <T2>}{<S2,S2>, <T2>}

Iteratively add links

t

Progression based approach

18

Example

Available behaviors

Target System

{<S1,S1>, <T1>}{<S2,S1>, <T1>}{<S2,S2>, <T1>}{<S1,S2>, <T1>}

{<S2,S1>, <T2>}{<S2,S2>, <T2>}

{<S2,S2>, <T3>}

Iteratively add links

X

Implementation

19

Implementation of both the techniques on a common platform

• Implement both approaches on a common platform – Java

• Prototype implementation available.

1. TLV implementation for deterministic available behaviors is available, but not for non-deterministic behaviors. Symfony is another system, but Lacks some of the components.

Analysis

20

Comparing the speed of the techniques

• Measure the speed of both the algorithms for the problems

• Design benchmark problems– Hand crafted• Problems for which a known solution exists• Problems for which a solution does not exist

– Randomly generated problems– Variation in size and number of available behaviors

• If time left in minor thesis– Study algorithm’s behavior with respect to• Varying degrees of non determinism in available behaviors

Questions ?

Comparing the speed of the techniques