CSI4124SYS5110 Foundations on Modeling and Simulation

1
CSI4124/SYS5110 Foundations on Modeling and
Simulation

• Discrete Event Simulation Modelling Event
  Scheduling Simulation Models

2
Topics of discussion

• Discrete Event Simulation Models 3 World Views
• Activity Scanning
• Event Scheduling
• Process-oriented
• Event Scheduling Simulation Models
• Principles of Event Scheduling Simulation Models
• Time Advance Algorithm
• Event Scheduling Simulation Models in Java
• Translating an ABCMod to an Java Event Scheduling
  Simulation Model
• Example Kojos Kitchen

3
Simulation Model World Views

• Activity Scanning
• ABCMod is base on this world view
• Implementation requires incrementing time in
  small steps, testing the pre-conditions of
  activities.
• Not efficient time advance algorithm
• Event Scheduling
• Expressed in terms of future events (events that
  follow some duration).
• Such future events include testing of the model
  state that triggers conditional events.
• Time is advanced when a future event is
  processed.
• Process-oriented
• Defines the processes (life-cycles) for entities
  (typically the consumer).
• ABCMod activities can be assembled into such
  processes.
• Time is advanced using scheduled events.

4
Activity Scanning Department Store
5
Event Scheduling Department Store
6
Process-oriented Department Store
7
Relationship between World Views

• Each world view is a perspective on the same set
  of discrete events
• Conditional and schedule events
• Relationships between events allow basis for
  translation from ABCMod to simulation models
• From ABCMod to event scheduling simulation Model
• Break down the Activity to events.
• Scheduled events become future events.
• Include conditional events as part of future
  events.
• From the ABCMod to a process-orient simulation
  model
• Build processes by assembling activities
• Requires an intermediate step augment
  conceptual model to define processes from
  activities

8
Kojos Kitchen
9
Event Scheduling Simulation Model

• Model behaviour
• Represented as a series of snapshots - change at
  discrete points in time, the future event.
• Snapshots contain state of the model (entities
  and attributes), list of scheduled future events,
  the simulation clock.
• Concerned with processing future events
• Future event encompasses both notions of ABCMod
  scheduled and conditional events.
• Composed of sequence of actions that changes the
  snapshot.
• Begins with the schedule event.
• Checks on model state to process any triggered
  conditional events.
• Processing events can lead to scheduling
  additional future events.
• This view is fundamental to the translation of
  the ABCmod to an event scheduling simulation
  model.

10
Event Scheduling Simulation Model Programming
Mechanisms

• Scheduling Future Events
• Future Event List (FEL) composed of future event
  notices
• Future event notice
• Contains a time attribute time at which the
  future event occurs
• Contains a future event name primarily used to
  identify a future event routine (FER).
• May contain a reference to a model entity used by
  the FER.
• Future Event Routine (FER)
• Routine that carries out the actions associated
  to the future event.
• Can also schedule new future events by adding
  event notices on the FEL
• Bootstrapping future event that reschedules
  itself
• Time advance algorithm
• Advances time and updates the model state
• Achieved by processing the event notices on the
  FEL

11
Time Advance Algorithm

• Loops until stop event notice is encountered or
  stop condition becomes true.
• Stop event notice contains Stop future event
  name.
• Stop condition consists of evaluating the state
  of the model.
• Event notice provides
• Time attribute to advance the model clock
• A future event name that can be associated to a
  FER
• The FER provides the necessary actions to be
  carried out for the event.

12
Bootstrapping

• Means of handling arrivals and input functions
• Recall the input domain sequence.
• Future event that reschedules itself
• Example Customers arrival in Kojos Kitchen
• First arrival is scheduled at start of simulation
  run by placing an event notice on the FEL, say
  with name WArrival, for arrival of a sandwich
  customer.
• When notice is processed, another will be
  scheduled as follows
• Generate an inter-arrival time, say a, using the
  data model associated with the timing map.
• Establish the next arrival time as, t t a,
  this will correspond to the next time in the
  domain sequence CSDUW.
• Insert into the FEL a new event notice with name
  WArrival and time attribute t.
• Bootstrapping can also be used for supporting
  inputs
• Consider manufacturing machine that breaks down
  (cycles of breakdowns and uptime).
• Consider the arrival of storms in the port
  project.

13
FER Actions

• Carry out the changes to the models status
  associated with a scheduled event.
• Typically corresponds to SCS of Activitys
  terminating event.
• Check the various preconditions to determine if
  any conditional events can be activated.
• Testing the preconditions for all Activities
  (more efficient approaches are possible). For
  each precondition that is found to be TRUE, the
  FER
• carries out state changes associated with that
  conditional event (i.e. the SCSs of the
  corresponding Activitys starting event) and
• schedules a future event derived from the
  corresponding Activitys terminating event.
• Second step needs to be repeated until all
  preconditions are FALSE.

14
Example EndServing Future Event

• Consider the future event EndServing for Kojos
  Kitchen
• Scheduled when customer arrives at counter for
  service (occurs when service is completed).
• Up to 3 event notices can be schedule with event
  EndServing.
• EndServing FER does the following
• Remove customer from counterGroup (note that
  reference to customer is necessary in event
  notice).
• Room at counter (counterGroup) has been freed. If
  a customer is present in the customer queue, then
• Remove customer entity from the head of the
  queue, and place it at the counter.
• Insert the customer waiting time into the
  PHIWaitingTime sample set.
• Schedule another EndServing future event

15
Event Scheduling with Java

• Java Standard Objects
• Representing Entities
• The Java Collections Framework library provides
  many classes for creating queues and groups
• The Java Management Extensions (JMX) library
  provide classes for defining attribute-tuples.
• CERN Colt Java Library
• Provides a number of classes that implement
  stochastic data models
• Link http//dsd.lbl.gov/hoschek/colt
• EVSched Simulation Package
• Abstract class for implementing simulation model
• Implements the FEL, time advance algorithm, event
  notices, adding event notices on the FEL,
  implementing FERs, implementing Stop condition,
  simulation clock, attribute-tuples and output
  sets (trajectory sets and sample sets)

16
EvSched Abstract Class
17
Event Notice
18
// Run simuation public void runSimulation()
while(true) // set up loop EventNotice
nxtev (EventNotice) fel.poll() if(nxtev
null) // This is a safety check
System.out.println("FEL is empty -
terminating") break
clock nxtev.timeAttr // update the clock
if(nxtev.eventName StopEvent) // Check for
a stop event System.out.println("E
ncountered stop event - terminating")
break processEvent(nxtev.eventName,
nxtev.obj) // Call FER if(implicitStopCond
ition()) System.out.println("Implicit
stop is true -terminating") break
timef clock
19
ESAttributeList
20
ESOutputSet
21
Computing DSOVs in ESOutputSet
22
Translating ABCmod to EvSched Simulation Model

• Translating Entities to Java Objects
• Use of Java Collections Library for groups and
  queues
• Use of ESAttribute Class for attribute-tuples
  (resources and consumer entities).
• Can also used Java Classes directly
• Translating Activities to FERs
• Decomposing Activities to create futures events
• Example Kojos Kitchen

23
Relating ABCmod Components to an Event Scheduling
Simulation Model
24
Representing Queue Entities in Java
25
Representing Group Entities in Java
26
Translating Activities to Future Event Routines
27
Steps to Translate an ABCmod to an Event
Scheduling Simulation Model

• Step 1 Represent entities, constants,
  parameters, input variables, etc. using
  appropriate data structures.
• Step 2 Implement appropriate data models and
  user defined modules specified in the conceptual
  model and an initialisation routine (EvSched
  constructor).
• Step 3 Define a FER for each Activity within
  the ABCmod.
• For any particular Activity, implement the SCSs
  in the Activitys terminating event or its action
  event. (in the case of an action sequence). Call
  the precondition routine.
• In the case of action sequences implement the
  requirements of bootstrapping.
• Step 4 Implement a precondition routine to
  start Activities when their precondition is TRUE.
  
• Step 5 Develop the program code to generate the
  required output data.

28
Kojo Kitchen Simulation Model Step 1

• C.Customer
• ESAttributeList with attributes
• Type (Character object with values W or S)
• TimeEnterQu (double value)
• A.CounterGroup
• HashSet object counterGroup.
• A.CustQue
• ConcurrentLinkedQueue object custQue.

29
Kojo Kitchen Simulation Model Step 1 (continued)

• Constants and Parameters
• WMean1, WMean2, WMean3, WMean4, WMean5
• UMean1, UMean2, UMean3, UMean3, UMean5
• STWMin, STWMax, STMin, STMax
• NumEmpReg (2 regular number of employees),
  NumEmpBusy (3 number of employees during the
  busy period)
• AECase1 (1 indicating base case for executing the
  model), AECase2 (2 indicating alternate case for
  executing the model).
• AddEmpCase (parameter) An int variable set to
  AECase1 for the base case (only 2 employees
  serving at the counter) and AECase2 when an
  additional employee is added during busy periods.
• Input Variables
• EmpNum (input variable) An integer variable
  that represents the number of employees at the
  Kojos Kitchen.

30
Kojo Kitchen Simulation Model Step 2

• Four objects serve as data models
• sanwichInterArrDist (Exponential object)
• sushiInterArrDist (Exponential object)
• coldCutSrvTm (Uniform object)
• sushiSrvTm (Uniform object)
• Three User Modules
• getMUw implements MUW(t)
• getMUu implements MUU(t)
• getMEmpNum implements MEmpNum(t)
• Initialisation
• Implemented in the KojoKitchen constructor

31
Kojo Kitchen Simulation Model Step 3

• Step 3 Define the future events

32
Step 3 processEvent method
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
Kojo Kitchen Simulation Model Step 4

• Define method preConditions to test all Activity
  preconditions
• All FER methods will call this method
• Checks the pre-conditions of the ServingW and
  ServingU activities

37
(No Transcript)
38
Implementing Other Functionality

• Consider the Port Project studied previously
• Define the following future events

39
Implementing Triggered Activities
40
Implementing a User Defined Module
41
Recording Trajectory Set Data
42
Interrupts