Building a SelectiveFidelity Simulation Engine

1
Building a Selective-Fidelity Simulation Engine

• Bradley C. Schricker
• ATT Government Solutions, Inc.

2
Presentation Outline

• Introduction of the Problem
• Simulation Fidelity
• Selective-Fidelity Simulation
• Selective-Fidelity Simulation Engine
• Summary Conclusion
• Questions
• Author Information

3
Introduction

• Balance between resource availability and
  fidelity
• MS community has recognized this issue and
  provided ideas such as multi-resolution
  simulation (Agg-Disagg work in OneSAF Testbed)
• With Selective-Fidelity Simulation, the simulator
  itself would adjust its fidelity

4
Fidelity

• Seems to be an intuitive understanding of
  fidelity
• No standardized definition
• More importantly, no standardized metric for
  measurement
• Currently, statements such as, this model is
  higher fidelity than that model, have little
  meaning

5
Fidelity, cont

• For the purpose of this discussion, fidelity
  means
• A measurement of how well a simulation
  represents its corresponding referent, which is
  based on a real-world object or system

6
Fidelity, cont

• For the purpose of this discussion, fidelity
  means
• A measurement of how well a simulation
  represents its corresponding referent, which is
  based on a real-world object or system

7
Fidelity, cont

• What is a referent?
• A referent is an abstract representation of any
  real-world object or system

8
Fidelity, cont

• Why do we need a referent?
• Impossibility of making a direct comparison
  between a model and the real-world system it
  represents
• Example
• An automobile simulation
• Direct comparisons can only be made between
  models
• Must have a referent model and then all other
  models

9
Fidelity, cont
10
Selective-Fidelity Simulation

• Fidelity Evaluation Framework
• Conceived by R. Franceschini and B. Schricker at
  Institute for Simulation and Training
• Method for measuring simulation fidelity
  pseudo-objectively
• Subjectivity is isolated from actual evaluation

11
Selective-Fidelity, cont
12
Selective-Fidelity, cont

• Critical Issues
• Self-monitoring of resources

13
Selective-Fidelity, cont

• Critical Issues
• Self-monitoring of resources
• Model organization and execution

14
Selective-Fidelity, cont

• Critical Issues
• Self-monitoring of resources
• Model organization and execution
• Toggling of models

15
Selective-Fidelity, cont

• Critical Issues
• Self-monitoring of resources
• Model organization and execution
• Toggling of models
• Spikes and valleys in resource usage

16
Selective-Fidelity, cont

• Critical Issues
• Self-monitoring of resources
• Model organization and execution
• Toggling of models
• Spikes and valleys in resource usage
• How often to poll resource monitors

17
Selective-Fidelity, cont

• General Strategy
• Modify code of simulation systems
• Not portable
• Requires intimate knowledge of the individual
  simulation systems
• Create separate modules for functionality
• More portable
• Requires knowledge only of data coming from
  individual simulation systems

18
Selective-Fidelity, cont

• Either way, two main components must be
  implemented in one way or another
• Resource Monitor
• Model Controller

19
Selective-Fidelity, cont

• Resource Monitor
• Monitors resources, such as
• CPU load
• Memory load
• Graphics memory load
• Most straightforward method is to continually
  poll the OS for this data
• In constant contact with the Model Controller

20
Selective-Fidelity, cont

• Model Controller
• Selects models based upon data from the Resource
  Monitor
• In any simulation system, implementation for
  starting execution of a model must already exist
• Thus, it should be relatively straightforward to
  reuse this to stop a model and start a new one

21
Selective-Fidelity, cont
22
Selective-Fidelity, cont
23
Selective-Fidelity Simulation Engine (SFSE)

• Houses the aforementioned functionality that
  applies to Selective-Fidelity Simulation
• Has similar functionality to High Level
  Architecture Runtime Infrastracture
• Must have knowledge of different simulators
  (which represent different models)
• Subscription and publication
• Ownership management

24
SFSE, cont
25
SFSE, cont
26
SFSE, cont

• Registration
• Publication
• Subscription
• Ownership Notification

27
Summary Conclusion

• Fidelity
• Fidelity Evaluation Framework
• Basic Selective-Fidelity Simulation Concept
• SFSE Concept
• Registration, Publication, Subscription, and
  Ownership Notification

28
Summary Conclusion, cont

• Future Research Ideas
• Terrain representations
• Automatically increasing fidelity of local
  terrain representations while decreasing fidelity
  of distant ones
• Ph.D. Dissertation
• Implementing a working SFSE and, through rigorous
  scientific testing, measuring the results

29
(No Transcript)
30
Author Information

• Bradley C. Schricker
• ATT Government Solutions, Inc.
• 11301 Corporate Blvd. Suite 110
• Orlando, FL 32817
• [phone]
• bschricker_at_att.com