| Notes |
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
Sentient World Simulation (SWS):
A Continuously Running Model of the Real World
A Concept Paper for Comments
Government POC
Tony Cerri
Anthony.Cerri@je.jfcom.mil
JFCOM J9,
Experimentation Engineering Lead,
757-203-3184
FAX 757-203-3198
Technical POC
Dr. Alok Chaturvedi
alok@purdue.edu
Purdue University
West Lafayette, IN 47907
765-494-9048
Version 2.0
August 22, 2006
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
Introduction
Modeling and simulation quickly becomes out of sync with new events, the emergence of new forces, and
newly proposed theories. The goal of the Sentient World Simulation (SWS) is to build a synthetic mirror of
the real world with automated continuous calibration with respect to current real-world information, such as
major events, opinion polls, demographic statistics, economic reports, and shifts in trends. The ability of a
synthetic model of the real world to sense, adapt, and react to real events distinguishes SWS from the
traditional approach of constructing a simulation to illustrate a phenomena. Behaviors emerge in the SWS
mirror world and are observed much as they are observed in the real world. Basing the synthetic world in
theory in a manner that is unbiased to specific outcomes offers a unique environment in which to develop,
test, and prove new perspectives.
SWS consists of components capable of capturing new events as they occur anywhere in the world, focus
on any local area of the synthetic world offers sufficient detail. In other words, the set of models that make
up the synthetic environment encompass the behavior of individuals, organizations, institutions,
infrastructures and geographies while simultaneously capturing the trends emerging from the interaction
among entities as well as between entities and the environment. The multi-granularity detail provides a
means for inserting new models of any temporal and spatial scales, or for incorporating user-supplied data
at any level of granularity. Therefore, SWS can be continuously enriched and refined as new information
becomes available.
SWS consists of the following components:
• A synthetic environment that supports Effects Based Approach and a comprehensive
representation of the real world at all levels of granularity in terms of a Political, Military,
Economic, Social, Informational, and Infrastructure (PMESII) framework.
• A scalable means of integrating heterogeneous components across time and space granularities.
• Mechanisms that discover, gather, and incorporate new knowledge into the continuously running
synthetic environment.
• A single façade of user interfaces enabling information from all sources (simulation generated
data, parameters for models, and data gathered from the real world) to be searched, viewed and
modified in an ontology-aware manner.
• Integrated Development Environments (IDE)s for constructing and configuring new models or
modifying existing models and then incorporating these changes into the continuously running
synthetic world.
• A means to take excursions from any point in time in the synthetic world to focus on select
regions of the world, leverage private user data, or to research specific theories by simplifying the
types of models to employ in the excursion.
SWS Components
The Core Synthetic Environment That Supports Pluralism of Thought
The core component of SWS is an agent-based environment named the Synthetic Environment for Analysis
and Simulation (SEAS), designed to be agnostic to the type of simulations and choice of models in order to
allow experimentation in the context of multiple and potentially conflicting theories. No single theory can
adequately explain complex behavior, such as the rise or splitting of terrorist organizations. Each theory
brings another perspective to the same phenomena. Only by combining these theories within the same
environment can we gain a comprehensive perspective.
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
Live Input &
Knowledge
Discovery
Excursion Manager
Analysis
Visualizations
SWS
Reference
World
Analyst
Interfaces
& Tools
Excursions
On Ramp
Off Ramp
Real World
Web Spiders
Live Sensors
Clustering
Semantic Extraction
Ontology -Based
Data Integration
Analytics
Model Parameterization
Model Development
Toolkits
Model Development
Toolkits
Synthetic World
Ontological
Representation
PMESII Models
Continuous Refinement
Models & Data
No Longer Needed
Cross -Excursion Analysis
Configurable Excursions
Collects Initialization Data
Selects Subset of Models
Persists Results
Copies of Ref . World
Use Private Data
Branch from
Calendar Date
Figure 1: A conceptual view of the Sentient World Simulation
By basing the designing of the synthetic environment on a fractal representation, SWS avoids the most
common pitfall faced by simulation environments: designing and optimizing an infrastructure for a
parochial purpose and thereby limiting its application to a small range of problems. The fractal approach
provides a consistent representation of the environment’s building blocks to models with all granularities of
access. The basic building blocks consist of individuals, organizations, institutions, information, and
geographies. From these building blocks, communities are built, constructed into societies, nations, and
ultimately providing a synthetic world. Coarse-grained models use a small sample of the population and
localized fine-grained models use a full sample of the population in an area. The same individual who
experiences events from a fine-grained model can be influenced by a coarse-grained model. This approach
to enabling heterogeneous models to interact is a Society of Models approach, described below.
SEAS provides all aspects of behavior within the PMESII categorization and enables the application of
Effects Based Thinking to the reference world of SWS.
A Society Approach to Integration
The cross-disciplinary nature of SWS fosters collaboration among diverse modeling and simulation
developers. This diversity necessitates an approach to integration that does not limit developers of SWS
components to adopt a uniform design, especially in terms of the temporal and spatial granularities with
which the components interact with the synthetic environment. The Society Approach implemented in the
SimBridge technology addresses the need to integrate components in a scalable manner while preserving
heterogeneity.
The Society Approach is applied to three areas in SWS to integrate models, simulations, and components.
A Society of Models enables models to operate at diverse temporal and spatial granularities within the
same synthetic environment. A Society of Models is employed by the Synthetic Environment for Analysis
and Simulation (SEAS) to integrate models ranging from global, regional, and national PMESII behaviors
to near real-time emotional arousal.
A Society of Simulations captures the necessary data exchange among interacting simulations without
requiring unscalable and difficult to maintain, centralized management. A Society of Simulations has been
used to integrate SEAS with a tactical, kinetics federation of simulations, Joint Semi-Automated Forces
(JSAF). This integration is described more fully in Figure 1.
A Society of Systems extends the Society of Simulations to include non-simulation components, such as
third party visualizations, enterprise system databases, analysis tools, and live sensor feeds.
A Society Approach fulfills the need of integrating with the goals of:
• independent and concurrent development of components,
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
• heterogeneous and reusable components,
• and high fidelity of detail in a scalable system.
Sensitivity to Current Information Through Real-Time Knowledge Discovery
Continuously incorporating current information in SWS provides three key benefits that distinguish SWS
from other initiatives at constructing a simulated environment:
• SWS remains up to date with respect to events and emerging trends.
• SWS leverages the prodigious amounts of data from all publicly available data sources,
something that is infeasible for a small number of analysts to gather in a timely manner.
• Models used by SWS are continuously refined, parameterized, and validated, keeping the
underlying model base of SWS relevant across time.
Burgeoning technology in the area of knowledge discovery has matured so that Web crawlers and spiders
are now used in research and industry. Applying this technology to news portals, blogs, and other internet
sources enables large amounts of data to be gathered and processed in a short amount of time.
By considering all available data, automated data mining provides an unbiased means of incorporating data
originating from multiple sources, and therefore, data from multiple perspectives. Additionally, interesting
outliers are discovered through text, video, and transaction analytics.
Believability and reliability metrics are applied to weight the influence of data from different sources
depending on the type of source, experience with data from the source, and the type of data. The
believability and reliability are then taken into account when incorporating the data into the SWS synthetic
world.
The discovery technology is coupled with a semantic engine that extracts semantics from the data. The
semantics are used to prepare the gathered data for use by the simulations and to relate the data to
knowledge already in the synthetic world.
Other sources of data besides the internet are incorporated using the same knowledge discover and
semantic extraction, such as proprietary, enterprise system, and classified data sources (classified and
proprietary data would only be incorporated in select excursions.)
The Ontological Representation of Knowledge
Using an ontological repository of knowledge, SWS augments analyst knowledge with simulation
semantics. The SWS ontological representation differs from traditional database approaches in that it stores
ontology as well as data. Ontology is a methodology for categorizing and annotating data based on logical,
human conceptualizations of what the data means. Interactions with the ontological annotations of data in
the repository are fundamentally different from the interactions allowed by the prevailing RDMS and
OODBMS approaches to database management. Traditional approaches coordinate tools with data by
identifying the physical data in queries whereas SWS enables coordination based on the semantic meaning
of data. A piece of data can be annotated with different keywords from diverse disciplines. Analysts as well
as automated processes can describe data needs using keywords from their own domains, enabling crossdisciplinary integration and a unified view of both real and synthetic information. A non-ambiguous data
categorization and annotation process is also needed to support the distributed and incremental process.
Data is also tagged with its informational sources and temporal lifetime. Consequently, multiple
experiments can be performed utilizing the same models combined with different points of views by
sourcing data from different informational sources or conducting an experiment in the context of a specific
calendar date.
A seminal integration of a knowledge repository with an active simulation environment was successfully
deployed in Urban Resolve 2015 (UR 2015). The implementation of an ontological representation of
knowledge, called eXtensible Net Assessment (xNA), was used to augment analyst knowledge with
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
simulation semantics. Including xNA as a component of the prototype SWS deployed in UR2015 extends
the Service Orientated Architecture (SOA) approach to provide a semantics-based integration of analyst
knowledge, stored in a repository named Operational Net Assessment (ONA), with emergent simulation
data. xNA supports multi-granularity and multi-disciplinary data to facilitate integration in the context of a
diversity of models and data sources. xNA can also receive incremental data updates from analysts and
automated data collection tools. More details on the prototype SWS are given in “JFCOM J9 Deployment
of SWS 0.5” below.
Model Development IDEs
SWS provides a set of tools based on the ontological repository for use by experts from various domains to
develop and experiment with models in the synthetic environment. A researcher begins with a set of
theories to test. Using algorithms of how the theories interact with the rest of the synthetic world, the
researcher creates models. The researcher designs a complete experiment environment using the new
models and leveraging existing models by creating a profile consisting of a selection of parameters from
the ontological data repository that fulfill the input requirements and support the experiment’s scenario.
SEAS Integrated Development Environment (SIDE) provides the tools to create and retrofit the artificial
agents in SEAS. Researchers will first design new types of agents along with the DNA and memory these
types of agents will have using the design interfaces of SIDE. The design interfaces also allow the DNA
and memory of existing agent types to be modified. After creating new agent types, SIDE is used to
configure the population makeup of the new agents in the synthetic environment.
After the creation of new agents, researchers use Just-in-time Modeling Environment (JIME) to create and
modify the behaviors of agents. Agent behaviors can be mathematical models based on variables that the
agents sense from the environment or can be described procedurally using a workflow engine, a
customizable system composed of states, transitions, and messages. The behaviors describe how agents
interact with their environment and other fellow agents.
Before incorporating the new agents into the reference world of SWS, agents are prudently tested in the
Bullpen tool. Using Bullpen, the behavior of agents is observed in a limited and controlled environment.
Researchers create profiles of environment variables for Bullpen experiments to observe the range of
decisions an agent makes and agent’s resulting actions.
Additionally, the modeling environment of SWS is extensible using ModelNet. ModelNet is a collection of
web-services that provides the functionalities of the aforementioned tools. Using the web-services, external
modeling tools can be integrated into the SWS to enable researchers to use modeling tools that are best for
their domain.
Excursion Management
SWS provides a configurable interface for configuring an excursion from the continuously running SWS
reference world to meet the individual needs of a user. User needs addressed by the Excursion Manager
include:
• Exploring multiple courses of action by taking different sets of actions in identical copies of the
reference world.
• Using proprietary or classified data in a controlled experiment without interfering with the
publicly accessible SWS.
• Constructing a synthetic environment for only a portion of the world or including only certain
models, simulations, tools, visualizations, or data sources.
• Conducting simultaneous excursions in different areas of the world and merging the
nonproprietary and unclassified results together.
Once an excursion is configured, the Excursion Manager fulfills the following:
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
• Designs a Society of Simulations, consisting of the user-configured choice of simulations and
components and any other components that the simulations depend on (referred to as members.)
• Sets up shared reality, the space shared among all members that are active in the excursion.
• For each excursion, pulls appropriate information for a specified calendar date from the
ontological repository to meet the members’ input requirements.
• Persists significant simulation results in the ontological repository, tagged with this excursion’s
identifier, for later analysis and cross-excursion analysis.
JFCOM J9 Deployment of SWS 0.5 for UR 2015
JSAF Federation
FAARS
Kinetics
CultureSim
CBSim
TacSim
MARCI
ONA External Data
Sources Multiplayer HITL
Ontological Structure
Multi -COA Persistence
Dynamic Sim . Updates
Pull -Based Queries
Time Ranges
Past & Real -Time
Supports Multi -COA
SDS
xNA
Data Selection
COA Initialization
COA Persistence
COA Manager
Action Planner
IED Planner
Input Tools
Shared Reality
SimBridge
Insurgency , IED
Networks
Dynamic Infra . Tree
Emotional Arousal
Will-toFight
Multi -Granularity
SEAS
Glyph Graph Viz
Nexus Leader Viz
Analysis Tools
Figure 2: Components of SWS v0.5
JFCOM showcased the first prototype of SWS, SWS v0.5, in the Urban Resolve 2015 (UR 2015) exercise.
UR 2015 was set in a synthetic urban setting and consisted of both simulation and real-time HITL inputs.
SWS 0.5 consisted of SEAS environment, a Society of Simulations and Systems implemented through
SimBridge, a deployment of xNA with knowledge from a classified repository and data from research on
public data, and an Excursion Manager.
Synthetic Environment for Analysis and Simulation (SEAS)
JFCOM has used SEAS to conduct a number of HITL experiments, providing the ability to evaluate the
Effects Based Approach, in Sea Viking 05, Multinational Experiment 4, and UR 2015. Each deployment
built on the previous, serving as stepping stones in the development of SWS.
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
Virtual International System (SEAS-VIS)
The SEAS-VIS model is a representation of the Institutions, Organizations, Leaders, Individuals, and
Infrastructure that make up a society. The geography of a society is modeled at various levels including
City, Province, Country, Region, and World in terms of Political, Military, Economic, Social, Information,
and Infrastructure (PMESII) nodes. The virtual environment integrates multiple theories from various
disciplines to program behaviorally accurate agents with dynamic rules that govern and guide their actions
and interactions. SEAS-VIS facilitates a seamless and interchangeable integration of human and software
agents and allows for testing the efficacy of theories, decisions, strategies, and tools.
SEAS-VIS offers the ability to do course of action (COA) analysis, test hypotheses, learn through
experimentation, conduct effects-based operations and planning, and develop and refine Operational Net
Assessment (ONA). The product allows the user to evaluate and manage the challenges of international
crises.
Near Real Time (SEAS-NRT)
SEAS-NRT consists of human behavior models that are best suited to a short period of time (within hours),
such as crowding and rioting, escalation of emotional arousal, and terrorist personnel collaborating in
improvised explosive device attacks. Following the fractal approach to representing an environment, the
SEAS-NRT models operate on the same entities that SEAS-VIS models operate on. In this way, SEASNRT and SEAS-VIS are coupled together in a Society of Models. Individuals respond to the media and
world public opinion models in SEAS-VIS, placing the individual in a dynamic global context, while also
enabling the individuals to sense and react to local events in real time.
SimBridge
Model integration within SEAS and application of SEAS to the tactical, kinetic realm, have been facilitated
by SimBridge, an implementation of the Society Approach. Through SimBridge, SEAS has been integrated
with Joint Semi-Automated Forces (JSAF) federations that implement the High Level Architecture (HLA)
approach to simulation integration. SEAS connected to JSAF through a liaison that joined the JSAF
federation as a single federate. In this way, SEAS was enabled to interact with CultureSim, CBSim, realtime HITL players, and many other components. SEAS interacts with CultureSim to represent real-time
population movement and activity. SEAS with input from a Chemical Biological Simulation (CBSim)
capture the influence of chemical and biological dispersions on population behavior.
Supporting the Society of Systems approach, the SimBridge Data Service (SDS) provides a query interface
through which visualizations, analysis tools, and external data base systems can leverage data residing in
shared reality. For UR 2015, SDS provided a selection of statistics representing the state of SEAS models
at various granularities to an external data repository, Future After Action Review System (FAARS), and to
Tactical Simulation (TACSIM), modeling human intelligence gathering. TACSIM used statistics such as
public opinion of Blue (one side in the multiplayer war game) to determine the willingness of the
population in a city sector to provide intelligence reports.
eXtensible Net Assessment (xNA)
Underlying SEAS is an ontological representation of SEAS knowledge within the eXtensible Net
Assessment (xNA) component. A repository populated by analysts’ understanding of Operational Net
Assessment (ONA) is matched with corresponding entities in xNA, enabling SEAS to draw from classified
ONA data to conduct certain classified excursions in UR2015. Also, selected types of changes that occur
within the simulation run time are persisted back into xNA, creating a dynamic ONA.
xNA represents knowledge tagged with the perspective, source, and excursion where the knowledge
originates from. This enables analysis tools to perform cross-COA analysis, such as tracking the emergent
growth of an insurgent organization in multiple scenarios where different actions were taken against terror
cells.
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
Course of Action Manager (COA Manager)
Users interacted with SWS 0.5 in both real-time interactions as well as for faster-than-real-time analysis.
Eight SEAS environments were used simultaneously to investigate the use of different sets of action plans
and to provide an unclassified interface with a select subset of events. One of the environments was placed
in a Society of Simulations with JSAF to capture events in real-time.
A COA Manager was implemented to enable each of the environments to be initialized and to enable
simulation data to be persisted, tagged with a COA identifier. A specific environment could be restarted by
accessing saved data for the desired COA.
The implementation of COA Manager is a step towards an Excursion Manager needed to facilitate SWS
excursions.
Use Cases of SWS
Training
• SWS provides the context within which training excursions are done, presenting the training
community with the unprecedented ability to train in a live and comprehensive synthetic
environment that is validated by theory and up to date with the real world.
• For skill refinement, a trainee’s skills can be directly compared with an experienced individual’s
skills by immersing the trainee into the synthetic environment in the real-world context where the
event actually occurred and analyzing the ensuing reactions of the simulated environment.
• Skill acquisition is supported by the ability to illustrate the emergent effects achieved by certain
doctrinal theories, and the proper contexts in which to apply them.
• The experimentation platform provides trainers with the ability to store and recall training
experiments, start another training experiment where an earlier one left off, and compare results
from multiple experiments.
Analysis
• Analysts’ theories are continuously weighed against emergent phenomena in SWS. As SWS
progresses with new real world events, the divergence between proposed theories or models
indicates a need for a change in thinking. Likewise, the synthetic world may confirm the
relevance of theories.
• Cross-excursion analysis reveals the influence of “what if” excursions from a comprehensive
perspective.
• The synthetic reference world of SWS provides immediate access to a mix of details from both
the real and synthetic worlds which remain up to date with respect to real world data and include
details which are not easily obtainable in the real world, such as measurements of emotional
arousal for a group of people.
• The analysis capabilities of SWS also reveal relationships between entities that emerge over time.
Analysts can perform full strategic, operational, and tactical net assessment with access to a
dynamic description of the links between nodes and across nodes and resources.
Planning
• SWS provides tools to develop, reuse, and compose action plans into playbooks. A playbook
database enables planners to recall playbooks.
• Planning tools enable plans to be developed for temporally and spatially fine-grained actions as
well as long-term actions and to place combined plans into a single playbook.
DRAFT DRAFT DRAFT
DRAFT DRAFT DRAFT
• Additionally, SWS facilitates the integration of Interagency participation in computer assisted
events.
Operations
• The execution environment enables plans to be utilized in a scenario and decision making to be
injected.
• Augmenting real-time information with near real-time and faster than real-time simulations
allows RCC to develop and test multiple courses of action to anticipate and shape behaviors of
adversaries, neutrals, and partners.
• SWS will provide the ability to develop and assess DIME campaigns.
Experimentation
• Users of SWS can take excursions that are configurable for the user’s unique needs, such as
selectively including or excluding models, interfaces, visualizations, and data sources.
• Users concerned with a localized region of the world can construct an experiment that uses only a
subset of SWS.
• The results from multiple excursions can be merged into a single picture using SWS’ Excursion
Manager, providing cross-excursion analysis.
• A subsequent excursion can be instantiated using the results from a previous experiment.
• Researchers can construct experiments to evaluate new models or employ user-supplied data
without interfering with the SWS reference world. The new models can be tested and refined in
an experimentation environment.
Testing
• SWS provides an environment for testing Psychological Operations (PSYOP) and Civil Affairs
activities, capable of illustrating the impact of these activities on populations.
• Commercial users can construct experiments to use proprietary data in a controlled environment. |