Software Praktikum Projects

Software Praktikum Projects

Stefan Resmerita

Winter semester 2013

2 Software Praktikum - WS2013 (Stefan Resmerita)

Projects 1&2: Modeling and Simulation of Multi-Agent Systems


Overview of projects 1&2

• Prerequisite: Java • Goal: simulate Multi-Agent Systems where agents

traverse a graph according to given rules and each agent is modeled as a DAG annotated with node occupation times (timed DAG)

• Projects: 1. Generate random inputs and check the output 2. Execute a given traversal algorithm by discrete-event

simulation

• Use the Ptolemy II framework http://ptolemy.eecs.berkeley.edu/ptolemyII/

http://ptolemy.eecs.berkeley.edu/ptolemyII/


Software for Modeling and Simulation

• Extend software frameworks by adding new components

• Create models of systems in the extended framework

• Run simulations and obtain results


Multi-Agent Systems (MAS)

• Agent: an entity (e.g., object in the OO sense) which behaves depeding on its own goals and on the behaviors and goals of the other agents – Example: an airplane

• Environment: the medium in which the agents evolve and interact – Example: the airspace and the traffic control system

• Emergent behaviors: characteristics of the entire multi-agent system resulting from the individual agent behavior – Example: airspace congestion


Simulation of MAS

• The MAS challenge: determine individual agent behaviors which lead to desirable emergent behaviors

– Example: given a set of airplanes, determine the set of possible trajectories for each airplane such that collisions and traffic congestion are avoided

• Simulation is one way to address this challenge


Our MAS

• Environment: a finite set of resources plus a mechanism to prioritize access to resources

• Agent: a finite set of behaviors, where a behavior represents a finite sequence of resource occupations

• An agent is represented by a set of trajectories which specify the resources and their occupation times

• Main requirement on the emergent behavior: no two agents may occupy a resource at the same time


MAS Example 1

• Conflict detection

• Prioritization

• Conflict resolution


MAS Example 2


MAS Example 3: Airspace application


MAS Example 3: Agent models


Project requirements

• Random generation of agent models

• Random generation of prioritization

• Run discrete-event simulations to obtain the solution according to the formal specifications

• Check the outputs


Discrete Event Simulation

Introduction

14 Software Praktikum - WS2013 (Stefan Resmerita) Slide from Lee & Seshia

Discrete Event Systems

• A dynamical system whose evolution is governed by the occurrence of events at discrete time points, at possibly irregularly-spaced intervals (Informal defn)

• Many cyber-physical systems are modeled as discrete-event systems: – Communication networks – Microprocessors – Manufacturing facilities – Communicating robots


Example: Communicating Robots/Sensor Nodes

Network can fwd, corrupt, drop packets

send recv


Simulating a Discrete-Event System (DES)

Discrete-Event System

Input Process

3. How to check Output?

2. How to generate random input?

i/p o/p

Environment

1. How to simulate the system on an input?


Simulating the System with an Event Queue

• Simulation Timer, T = 0

• Repeat while there are events in the event queue: 1. Dequeue event at head of queue (“imminent event”)

2. Advance simulation timer to time of imminent event

3. Execute imminent event: update system state

4. Generate future events and enqueue them

. . .

Event queue

t1

e1

t2

e2

t3

e3

timestamp

event record t1 < t2 < t3 < …


Example of Simulation with Event Queue

Network

1

2

3

e1 = send(1, 2, 00)

T = 0

. . .

Event queue

1.5

e1

e2 = send(3, 1, 10)

3.8

e2



Network

1

2

3

e3 = fwd(2, 1, 00)

T = 1.5

. . .

Event queue

1.6

e3

e2 = send(3, 1, 10)

3.8

e2



Network

1

2

3

e4 = recv(2, 1, 00)

T = 1.6

. . .

Event queue

3.8

e2

e2 = send(3, 1, 10)

4.1

e4


Implementing the Event Queue

• Event with smallest time-stamp must be dequeued

• New events must be inserted into sorted order according to their timestamps

• Efficient Data Structure: Priority Queue

• Particular version: Calendar Queue

[R. Brown, Comm. of the ACM, 1988, vol. 31(10)]


Simulating a Discrete-Event System (DES)

Discrete-Event System

Input Process

3. How to check Output?

2. How to generate random input?

i/p o/p

Environment

1. How to simulate the system?

Input Process


Actor-Oriented Design and The Ptolemy II

framework http://ptolemy.eecs.berkeley.edu/


Ptolemy II objectives

• Supports modeling, simulation and design of concurrent systems

• Promotes component-based modeling, where a component represents a domain-specific entity and it is called an actor

• Provides widely-used models of interaction between components, called models of computation

• Advocates a programming discipline called Actor-Oriented Programming

• Focuses on flexibility • Encourages experimentation with designs


Actor Oriented Design

• Actors are conceptually concurrent (no predefined order of execution)

• Actors interact by sending messages through channels

• An actor implements an execution interface

• Dedicated components are responsible for data transfer between actors and for executing the actors

• Actors can be hierarchically composed


AOD versus OOD (I)

class name

data

methods

call return

What flows through an object is sequential

control

Object orientation:

Actor orientation:

actor name

data (state)

ports Input data

parameters Output data

What flows through an object is streams

of data


AOD versus OOD (II)

Identified limitations of object orientation: Says little or nothing about concurrency and time Concurrency typically expressed with threads, monitors, semaphores Components tend to implement low-level communication protocols

Re-use potential is disappointing

OO interface definition gives procedures

that have to be invoked in an order not

specified as part of the interface definition.

TextToSpeech

initialize(): void

notify(): void

isReady(): boolean

getSpeech(): double[]

actor-oriented interface definition says

“Give me text and I’ll give you speech”

Actor oriented Object oriented


Next project steps

• Read and understand the formal MAS description (to be sent by e-mail)

• Create and run simple DE models in Ptolemy II

• Links:

– The Ptolemy II web site: http://ptolemy.eecs.berkeley.edu/ptolemyII/

– The Ptolemy II book:

http://ptolemy.org/systems

http://ptolemy.eecs.berkeley.edu/ptolemyII/



Software Praktikum Projects

Documents