-
AGILE INFORMATION CONTROL ENVIRONMENT PROGRAM
E
Agile Information Control Environment Program
I-Jeng Wang and Steven D. Jones
fficient and precise management and control of information flow
are keycomponents in winning information warfare. We discuss two
new APL AgileInformation Control Environment Program projects for
the Defense AdvancedResearch Projects Agency. These two projects
will develop a theoretical foundationand enabling technology to
dynamically manage and control information flow across adistributed
and heterogeneous network infrastructure comprising military and
commer-cial networks. (Keywords: Adaptive information control,
Discrete event systems,Dynamic channel building, Global QoS
optimization, Tactical networks.)
INTRODUCTIONThe goal of the Agile Information Control
Environ-
ment (AICE) Program, established by the DefenseAdvanced Research
Projects Agency (DARPA), is todevelop a theoretical foundation and
enabling technol-ogy to dynamically manage and control
informationflow across a distributed and heterogeneous
networkinfrastructure. Future Joint warfare will require a
hugeamount of information sharing among several usersover many
dissimilar tactical and commercial networks.Depending on the
associated applications or missions,different information flows
might require differentquality-of-service (QoS) guarantees from the
networksthat carry them. Some possible QoS factors arethroughput,
maximum delay, reliability, and covertness.Several DoD programs
have been established to devel-op tactical network infrastructures
to facilitate informa-tion sharing with heterogeneous QoS
requirementsamong multiple sensory systems and military users,
forexample, the Cooperative Engagement Capability(CEC).
Nevertheless, no mechanism currently exists
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 3 (1
that permits dynamic control and optimization of in-formation
flow over multiple tactical and commercialnetworks at a scale
required by future Joint warfare. Theultimate objective of the AICE
Program is to achieveinformation control in a faster, more
efficient, and moreprecise way than our adversaries.
Four-Layer Functional Architecture
DARPA envisions a four-layer functional architec-ture for AICE,
as shown in Fig. 1.
• Physical network: This layer will consist of manyindependently
owned and operated tactical and com-mercial networks that in
general provide routingservices and unique QoS capabilities. These
are theactual communications resources on which the infor-mation
flows will travel.
• MetaNet: Future military information operationswill use a
variety of commercial and DoD communi-cations systems and networks.
Information will be
999) 271
-
I-J. WANG AND S. D. JONES
Information policymanagement
Adaptive informationcontrol
MetaNet
ACN Commercial GBS
DISN CEC
APLAICE
efforts
AICE
Physicalnetworks
Informationflow
Informationusers
Informationsources
Informationflow
Commanders
Figure 1. DARPA AICE functional architecture. (ACN =
AirborneCommunication Node, DISN = Defense Information System
Net-work, CEC = Cooperative Engagement Capability, GBS =
GlobalBroadcast Service.)
disseminated based on the QoS required by the user,and the
requisite networks will be selected on a per-message basis for
their ability to achieve QoS. TheMetaNet system facilitates
QoS-based routing overthis collection of networks. MetaNet has four
aspects:(1) inserting QoS-like capabilities into existing tacti-cal
networks to enable dynamic reallocation of net-work resources, (2)
negotiating service requests as anintermediary between the user and
individual net-works, (3) providing end-to-end QoS solutions
withintime constraints, and (4) maintaining negotiatedend-to-end
QoS by dynamically re-routing or renego-tiating service.
• Adaptive information control (AIC): This layer pro-vides
global “content-aware” dynamic informationflow control, using the
services of the MetaNet layerto do so. Features of the AIC layer
include partition-ing of information flows among available logical
chan-nels, globally optimizing allocation to achieve infor-mation
flow priorities for military users, developingprofiles of user
information dissemination needs, real-locating resources when
necessary owing to networkQoS degradation, and managing flow
control toaccommodate real-time communications.
• Information policy management (IPM): This layerhas three
primary functions: (1) it provides users thecapability to visualize
the effects of their informationcontrol policies, (2) it relates
information policymanagement to military operations, and (3) it
aidsin the synthesis of effective information controlpolicies.
272 JOHN
The DARPA AICE Program focuses on MetaNet, AIC,and IPM. APL’s
AICE efforts will include two 3-yearprojects beginning in January
1999 (the second andthird years are optional) focused on the
MetaNet andAIC layers. These two projects are collaborative
effortsinvolving APL, Purdue University, and the Universityof
Missouri-Columbia.
APL METANET PROJECTThe goal of the MetaNet Project is to develop
an
intelligent MetaNet controller (IMC). The main tasksof the
MetaNet layer are to negotiate and coordinatethe setup of
end-to-end QoS across multiple, dissimilarQoS networks and to
perform internetwork routing. Itwill provide a set of end-to-end
logical channels (eachlogical channel normally runs across several
physicalnetworks) with specific guaranteed QoS to the AIClayer for
information flow control. It will also monitorthe network traffic
status of each established logicalchannel and report any change in
channel QoS to theAIC layer for appropriate responses or necessary
infor-mation channel reconfigurations.
APL will develop an IMC as a component of theAICE Program. A key
feature of our design is the treat-ment of the process of delivery
of QoS to communica-tions system users in a control-theoretic
manner usingsupervisory control. The core controller will be
synthe-sized as a Discrete Event System (DES) model andsupplemented
by artificial intelligence and heuristictechniques. Unanticipated
events will be handled asperturbations to achieve a control state
defined by theQoS goal. The output of the controller will consist
ofactions necessary to overcome the perturbations.
The IMC will serve as an interface to four militarytactical
communications systems: (1) the Mobile Sub-scriber Equipment and
its Warfighter InformationNetwork-Terrestrial (WIN-T) follow-on;
(2) the JointComposite Tracking Network, which is a follow-on toCEC
and its Data Distribution System (DDS) commu-nications component;
(3) the Milstar Extremely HighFrequency and Advanced Extremely High
Frequencymilitary satellite communications (Milsatcom) system;and
(4) the Airborne Communications Node System.Access to these systems
will be controlled according tothe required QoS in a manner
transparent to the user.The IMC concept will allow more efficient
use ofavailable networks for tactical communications appli-cations
than is currently achieved by eliminating theuse of stovepipes in
which information is always sentvia a fixed medium.
The concept of operations for the IMC includesinterfaces to a
higher AIC layer and subordinate tac-tical communications networks.
The AIC will pass tothe IMC user requests for communications access
alongwith a set of QoS requirements. The IMC will assess
S HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 3 (1999)
-
the available connectivity and QoS through negotia-tion with,
and analysis of, the subordinate system’sperformance and capacity.
Resulting solutions will bepassed to the AIC, where connectivity
decisions aremade. The AIC will then inform the IMC, and theIMC
will direct resource allocation in the chosen com-munications
system to establish connectivity.
Military communications pose unique control chal-lenges because
of widely varying traffic priorities, rap-idly changing
environmental conditions, mobile net-worked field terminals, and
enemy interference.Current control technology uses very simple and
oftenad hoc algorithms to react to these conditions and failsto
exploit the richness of available software capabilities.
The IMC will build on concepts previously devel-oped by APL for
an intelligent controller for Milsatcomterminals as part of the
DARPA Program for IMPACT(insertion into Milsatcom products advanced
commu-nications technology). Sophisticated models ofnetwork
behavior can enable much-needed effectiveautomated diagnosis and
treatment of network QoSfailures. Available capacity cannot be
exploitedwithout using adaptive congestion management algo-rithms
that monitor and react tochanging traffic patterns.
Recentartificial intelligence research1,2 inthe area of
reinforcement learninghas made possible the design ofadaptive
scheduling and routing al-gorithms that react to traffic pat-terns,
enabling more efficient use ofbandwidth. Recent work3,4 in
su-pervisory control theory providesan additional perspective,
allowingthe use of adaptive tuning of pa-rameterized network
configura-tions to maintain QoS. A mathe-matical model of the
networkconfiguration management func-tions used in the IMC will be
cre-ated and evaluated for speed andaccuracy in assessing QoS,
firstwith models of the subordinatecommunications systems and
sec-ond in interface trials with thosesystems (see Fig. 2).
APL AIC PROJECTThe goal of the AIC Project is
to develop novel AIC techniquesfor the AICE Program. The
AIClayer will provide global content-aware dynamic information
flowcontrol using the content-unawareservices of MetaNet. It must
take
Figure 2 . Intelligent MeGUI = graphical user inMSE = Mobile
SubscribJCTN = Joint CompositACN = Airborne Comm
Assessmentmanager
AIheurisGUI
MSE/WIN-T
Dataacquisition
Physicaldata interface
QoSmodeling
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 3 (19
AGILE INFORMATION CONTROL ENVIRONMENT PROGRAM
into account the value of the information flow to thecommander
and optimize the use of overall QoS re-sources furnished by
MetaNet. Our approach is built ona set of basic functional
components, each performinga crucial subtask required by the AIC.
To respond toa rapidly changing dynamic environment and to
meetreal-time performance constraints, the AIC layer con-stantly
iterates over subtasks to accomplish the overallinformation control
objective. Figure 3 shows the fourfunctional components (AIC
subtasks) of the APLapproach:
1. Information flow aggregation. To enable effectivelogical
channel negotiation with MetaNet and quickreconfiguration in
response to various changes, theAIC layer first organizes the large
set of informationflows into a collection of smaller and more
manage-able groups by appropriate aggregation.
2. Logical channel negotiation with MetaNet. For eachgroup of
information flow in the established structure,the AIC layer
negotiates with the MetaNet to estab-lish one or more end-to-end
logical channels to satisfyits total QoS requirements.
taNet controller and interfaces. (AIC = adaptive information
control,terface, AI = artificial intelligence, DES = Discrete Event
System,er Equipment, WIN-T = Warfighter Information
Network-Terrestrial,e Tracking Network, AEHF = Advanced Extremely
High Frequency,unication Node.)
Dataacquisition
Physicaldata interface
QoSmodeling
Dataacquisition
Physicaldata interface
QoSmodeling
Dataacquisition
Physicaldata interface
QoSmodeling
Assessmentmanager
Assessmentmanager
Assessmentmanager
Requirementprocessing
ticsDES
engine
AIC layer
JCTN Milstar/AEHF ACNSystem/network
99) 273
-
3
4
mitimacE(
I-J. WANG AND S. D. JONES
. Channel assignment to information flows. Based onthe set of
end-to-end logical channels from MetaNet,which may not be exactly
what was originally re-quested, the AIC layer assigns to each
informationflow a logical channel that meets the QoS require-ments.
In general, more than one information flow isassigned to each
logical channel.
. Global QoS optimization to allocate channel QoSresources. When
traffic congestion occurs as a resultof either channel degradation
or over-subscriptionduring the channel building stage, the AIC will
opti-mize the allocation of overall channel QoS resourcesamong
information flows in accordance with com-manders’ operational
priorities.
In a dynamic environment where the required infor-ation flows,
network status, and commanders’ prior-
ties can all change with time, the AIC layer will needo iterate
over the described subtasks with various scalesn different orders.
To react to rapid changes in infor-ation control requirements, APL
will develop a scal-
ble incremental reconfiguration scheme that properlyoordinates
the implementation of the four subtasks.ventually, changes may
cause the AIC to recalculatein the background) a new set of
proposed logical
Figure 3. Functional components of AIC.
Flowaggregation
Channelnegotiation
Channelassignment
GlobalQoS
optimization
Dynamic channel building
274 JO
channels (i.e., subtasks 1 to 3 are repeated). At theconceptual
or abstract level, the AIC layer faces alarge-scale dynamic
resource allocation problem (deal-ing with 10,000 users and 1,000
sources simultaneously,as estimated by DARPA). The size of the
problem, thereal-time constraints, and the variety of
performancecriteria will easily render any exact optimal
solutionimpractical. We will develop efficient
approximatealgorithms that achieve near-optimal performance.
To better understand the issues and trade-offs pre-sented by the
large-scale adaptive information controlproblems faced by the AIC
layer, we will also developa mathematical framework that models the
relevantfactors and desired optimization criteria. This frame-work
will serve as a scientific basis for our developmentof efficient
AIC techniques for the AICE Program.Moreover, the framework will
provide a rigorous wayto assess the performance of techniques
developedthroughout the project. Finally, we will develop a
work-ing prototype for the AIC layer and demonstrate itsperformance
by simulating information flow control ina large-scale military
operation.
CONCLUSIONWith APL’s extensive experience in tactical net-
work management and control, as well as expertise inlarge-scale
computations and optimization, we believethat the two AICE projects
presented in this article willlead to innovative and exciting
technologies for theefficient dynamic information control of future
large-scale Joint warfare.
REFERENCES1 Bertsekas, D. P., and Tsitsiklis, J. N.,
Neuro-Dynamic Programming, Athena
Scientific, Belmont, MA (1996).2Sutton, R., and Barto, A.,
Reinforcement Learning: An Introduction, MIT Press,
Cambridge, MA (1998).3 Ramage, P., and Wonham, W., “The Control
of Discrete Event Systems,”
Proc. IEEE 77, 81–98 (1989).4 Cassandras, C. G., Discrete Event
Systems: Modeling and Performance Analysis,
Richard D. Irwin, Inc., and Aksen Associates, Inc., Boston, MA
(1993).
ACKNOWLEDGMENTS: The authors wish to acknowledge the
contributionsof Howard Siegel and Edwin Chong of Purdue University
and Michael Jurczyk ofthe University of Missouri-Columbia, and
would also like to thank Frank Vaughanfor his assistance in
administrative work and program management The
principalinvestigators are Steven Jones for the MetaNet Project and
I-Jeng Wang for theAIC Project. This work was supported under
contracts DABT63-99-C-0010 andDABT63-99-C-0012 with DARPA.
HNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 3 (1999)
-
AGILE INFORMATION CONTROL ENVIRONMENT PROGRAM
THE AUTHORS
I-JENG WANG received an M.S. degree from Penn State University
in 1991and a Ph.D. degree from Purdue University in 1996, both in
electricalengineering. He joined APL in 1997 and is a member of the
System andInformation Sciences Group of the Milton S. Eisenhower
Research andTechnology Development Center. Dr. Wang’s research
interests include stochas-tic optimization, reasoning under
uncertainty, and large-scale network manage-ment and control. He is
a member of the Institute of Electrical and ElectronicsEngineers,
The Association for Computing Machinery, and the Institute
forOperations Research and the Management Sciences. His e-mail
address is [email protected].
STEVEN D. JONES received a B.S. degree from the University of
Maryland,College Park, in 1981 and an M.S. degree from The Johns
Hopkins Universityin 1985, both in electrical engineering. He
joined APL in 1981 and is currentlya member of the Principal
Professional Staff. He is the lead engineer for severalmilitary
communications programs with the Satellite Communications
SystemsGroup in the Power Projection Systems Department. His
experiences include theanalysis, simulation, and testing of
satellite communications links and equip-ment. Mr. Jones’s work has
largely involved the Milstar system, principally itsground terminal
segment. He is a member of the Institute of Electrical
andElectronics Engineers Communications Society and Tau Beta Pi.
His e-mailaddress is [email protected].
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 3 (1999)
275
mailto:[email protected]:[email protected]
