darpa initiatives - computing research association...
TRANSCRIPT
1
DARPA InitiativesDARPA Initiatives
Preston MarshallPreston MarshallDARPA ATO Program ManagerDARPA ATO Program Manager
–– neXtneXt Generation (XG) Communications Generation (XG) Communications ––
–– Connectionless Networking (CN) Connectionless Networking (CN) ––
–– Disruption Tolerant Networking (DTN) Disruption Tolerant Networking (DTN) ––
2
PurposePurpose
•• Describe Related DARPA ProgramsDescribe Related DARPA Programs–– neXtneXt Generation (XG) CommunicationsGeneration (XG) Communications–– Connectionless Networking (CN)Connectionless Networking (CN)–– Disruption Tolerant Networking (DTN)Disruption Tolerant Networking (DTN)
•• Present DARPA PerformersPresent DARPA Performers–– Potential Teaming PartnersPotential Teaming Partners
•• Describe Potential Synergy With Existing and Describe Potential Synergy With Existing and Planned EffortsPlanned Efforts
3
All Spectrum May Be Assigned, But…
…Most Spectrum Is Unused!
XG is Developing the Technology and System Concepts for DoD to Dynamically Access All Available
Spectrum
Frequency
Pow
er ReactFormulate Best
Course of Action
ReactReactFormulate Best Formulate Best
Course of ActionCourse of Action
AdaptTransition
network to new emission plan
AdaptAdaptTransition Transition
network to new network to new emission plan emission plan
CharacterizeRapid waveform
determination
CharacterizeCharacterizeRapid waveform Rapid waveform determinationdetermination
SenseReal time, Low-
power, wideband monitoring
SenseSenseReal time, LowReal time, Low--
power, wideband power, wideband monitoringmonitoring
AutonomousAutonomousDynamic Dynamic SpectrumSpectrumUtilizationUtilization
DARPA XG ProgramDARPA XG Program
Goal: Demonstrate Factor of 10 Increase in Spectrum Access
4
A New Way to Manage Spectrum Is NecessaryA New Way to Manage Spectrum Is Necessary
•• LargeLarge--Scale Ad Hoc Network Complexity Scale Ad Hoc Network Complexity Beyond Human PlanningBeyond Human Planning-- Dynamic topologiesDynamic topologies-- Planning is at least NPlanning is at least N22
•• Spectrum Regulatory Processes Vary Spectrum Regulatory Processes Vary Between Countries and RegionsBetween Countries and Regions-- Requires Extensive Coordination TimeRequires Extensive Coordination Time
Recent SUO Network Analysis
Secretary of State Telecommunications & Public Information
Secretary of State Secretary of State Telecommunications Telecommunications & Public Information& Public Information
Ministry of Communications
Ministry of Ministry of CommunicationsCommunications
Regulatory Authority of
Communications & Posts
Regulatory Regulatory Authority of Authority of
Communications Communications & Posts& Posts
Regulatory Authority of Telecommunications
and Posts
Regulatory Authority of Regulatory Authority of Telecommunications Telecommunications
and Postsand Posts
Ministry of Public Works, Transport,
& Water Mgt
Ministry of Public Ministry of Public Works, Transport, Works, Transport,
& Water Mgt& Water MgtDepartment of Trade & Industry
Department Department of Trade & of Trade & IndustryIndustry
Ministry of Science,
Technology, & Innovation
Ministry of Ministry of Science, Science,
Technology, & Technology, & InnovationInnovation
Ministry of Industry,
Employment, & Communications
Ministry of Ministry of Industry, Industry,
Employment, & Employment, & CommunicationsCommunications
Ministry of Transport &
Communications
Ministry of Ministry of Transport & Transport &
CommunicationsCommunications
Ministry of Civil Affairs &
Communications
Ministry of Civil Ministry of Civil Affairs & Affairs &
CommunicationsCommunications
Ministry of Economics, Finance, &
Industry
Ministry of Ministry of Economics, Economics, Finance, & Finance, &
IndustryIndustry
Ministry of Transport &
Communications
Ministry of Ministry of Transport & Transport &
CommunicationsCommunications
Ministry of InfrastructureMinistry of Ministry of
InfrastructureInfrastructure
Ministry of Transport, Posts, &
Telecommunications
Ministry of Transport, Ministry of Transport, Posts, & Posts, &
TelecommunicationsTelecommunicationsDepartment
of Public Enterprise
Department Department of Public of Public
EnterpriseEnterprise
5
The XG Problem SpaceHow Do We?
The XG Problem SpaceHow Do We?
Measure Instantaneous
Spectrum Usage?
Account for Propagation
Differences?
Policy LanguageCommon
Approaches that Must Be Agreed on, and Can be Adopted
WidelyResolve In
consistent
Policies?
Infer Ambiguous
Policies?
Protect “Hidden”
Nodes?
Reflect Band-Specific Policies?
Reflect Nation -
Policies?
Optimize System
Performance?
Infer Poten
tial
Interfer
ence
?
ImplementationDesign Specific
Approaches that Can be Implemented in
Many Ways to Develop Unique
Products
Describe Worst-Case Interference?
Abstract Capabilities Abstract Capabilities (Behaviors)(Behaviors)
6
XG Key PrinciplesXG Key Principles•• Suitable for Range of Architectural ImplementationsSuitable for Range of Architectural Implementations
–– From StandFrom Stand--alone to Fullyalone to Fully--cooperativecooperative–– Extensible Set of Control ProtocolsExtensible Set of Control Protocols–– Separate Policy Controls From Radio TechnologiesSeparate Policy Controls From Radio Technologies–– Implementation and Application (Military/Civil) IndependentImplementation and Application (Military/Civil) Independent
•• Identify “Identify “InterferenceInterference--PreservingPreserving” Core Set” Core Set–– Similar to SecuritySimilar to Security--Preserving FrameworksPreserving Frameworks–– Allow for XG Extension Outside of Core Set BoundaryAllow for XG Extension Outside of Core Set Boundary
•• Create Framework and Set of Abstract BehaviorsCreate Framework and Set of Abstract Behaviors–– Manage Communications Network Spectrum AccessManage Communications Network Spectrum Access–– Coordinate Other Spectrum Users (Radars,…)Coordinate Other Spectrum Users (Radars,…)–– Not Specific to Any One Physical or Link LayerNot Specific to Any One Physical or Link Layer
•• Protocol Developed & Managed Via Request For Comment (RFC)Protocol Developed & Managed Via Request For Comment (RFC)–– DARPADARPA--Run and Managed, but Open to Public for Comment/inputRun and Managed, but Open to Public for Comment/input–– Incorporate or Leave Hooks for Extensions of XG CapabilitiesIncorporate or Leave Hooks for Extensions of XG Capabilities
•• Enable Policy and Regulatory NeedsEnable Policy and Regulatory Needs–– Provide For Richness/Complexity of PoliciesProvide For Richness/Complexity of Policies–– Allow For Diversity of Policy SourcesAllow For Diversity of Policy Sources–– Facilitate Regulatory Traceability and AcceptanceFacilitate Regulatory Traceability and Acceptance
7
How to Technology and Policy Can Maximize Access
How to Technology and Policy Can Maximize Access
Three Ways to Make Opportunistic Sharing More Effective
1. Enhance Policy Flexibility by Opening Up the Envelope• Accept and Manage More
Risk
2. Increase Capability to Dynamically Sense and Adapt• Faster Spectrum
Analyzers, More Instantaneous Bandwidth
3. Develop Radios & Waveform Standards that Can Adapt to Meet Sharing Requirements
• Wider Coverage, Better Antennas, Adaptive Waveforms
Dim
ensi
on 2 Policy C
onstraint
Wav
efor
m C
onst
rain
t
OpArea
OpArea
Operating AreaOperating
Area
Dimension 1
DARPA Not Addressing These Challenges!
8
The FutureThe Future
•• Automated Spectrum Access Will HappenAutomated Spectrum Access Will Happen–– FCC Spectrum Task Force Recommendations on FCC Spectrum Task Force Recommendations on
Enabling Adaptive Spectrum Access TechnologiesEnabling Adaptive Spectrum Access Technologies–– Existing 802.11a ProtocolsExisting 802.11a Protocols
•• Sensing of holes and cooperative spectrum sharing Sensing of holes and cooperative spectrum sharing –– 4G Spectrum Management Technologies4G Spectrum Management Technologies
•• Immediate NeedsImmediate Needs–– Policy to Allow Initial Development and Usage Policy to Allow Initial Development and Usage
•• This Process, Task Force, …This Process, Task Force, …–– Spectrum Sensing and Adaptation Technology Spectrum Sensing and Adaptation Technology
•• DARPA XG, Others?DARPA XG, Others?•• Far Term (Not Current DARPA Programs)Far Term (Not Current DARPA Programs)
–– Radio Designs That Are Specific to the Adaptive Policies Radio Designs That Are Specific to the Adaptive Policies and Technologyand Technology•• Optimization, cost, spectrum leasing, operating Optimization, cost, spectrum leasing, operating
capabilities…capabilities…
9
XG ContractorsXG Contractors
•• Policy Language and ProtocolsPolicy Language and Protocols–– BBN TechnologiesBBN Technologies
•• System Integration and DesignSystem Integration and Design–– Lockheed Martin Advanced Technology LaboratoriesLockheed Martin Advanced Technology Laboratories–– RaytheonRaytheon–– Shared Spectrum CompanyShared Spectrum Company
•• Sensor DesignSensor Design–– Rockwell CollinsRockwell Collins
•• Advanced TechnologiesAdvanced Technologies–– Raytheon Raytheon –– Advanced WaveformAdvanced Waveform–– Shared Spectrum Shared Spectrum –– Transmitter Transmitter TechologiesTechologies
RockwellRockwellCollinsCollins
10
Connectionless NetworkingConnectionless Networking
•• AdAd--Hoc, Low Duty Cycle Networks Require a Different Kind of RadioHoc, Low Duty Cycle Networks Require a Different Kind of Radio–– Closed Loop Links and Networking Ineffective for Short Duty CyclClosed Loop Links and Networking Ineffective for Short Duty Cycle e
Messaging ApplicationsMessaging Applications–– Energy Required by RF Link and Network Topology Limits Lifespan,Energy Required by RF Link and Network Topology Limits Lifespan,
Miniaturization…Miniaturization…–– Constant Energy per BitConstant Energy per Bit
•• OpenOpen--Loop Potential Savings 10Loop Potential Savings 1044+ in + in Energy/Time or Increase in “Energy/Time or Increase in “GoodPutGoodPut””–– PrePre--Compute and Correct Waveform Compute and Correct Waveform
Frequency, Timing, and Phase for “Instant”Frequency, Timing, and Phase for “Instant”Acquisition/SynchronizationAcquisition/Synchronization
–– Processor Intensive Rather Than Data Processor Intensive Rather Than Data Intensive RoutingIntensive Routing
–– Availability of State (Relative Position & Velocity) and TimeAvailability of State (Relative Position & Velocity) and Time–– Size/Weight/Power of Sensing and RF Components Continue to DropSize/Weight/Power of Sensing and RF Components Continue to Drop
ChipChip--ScaleScaleSensorSensor
ChipChip--Scale RadioScale Radio
The Future With Current Approaches
Challenge The Assumption That Wireless Systems Should Look Like Wired Ones
Challenge The Assumption That Wireless Systems Should Look Like Wired Ones
11
Link Formation is Costly!Link Formation is Costly!• Even SUOSAS Takes Up to 10
Sec. to Complete Layer 1 to 3– MILSTAR Can Take Minutes
• End to End Acknowledgement Greatly Increases Delay
• Efficiency <0.1% For Short Messages
– LPI, LPD, Battery, Complexity Impacts
RF
Recent SUO Network Analysis
Managing This Network Comes at a High Cost
Significant Power/BW Spent on Overhead … Invest in Getting Efficiency from .1% to 10%
Significant Power/BW Spent on Overhead … Invest in Getting Efficiency from .1% to 10%
PHYSICAL
Resolve Driven By:Hop Time Distance/Time ToleranceFrequency Relative VelocityTiming Distance/Time TolerancePhase Baseband Coherent TrackingClock Rate Relative Closing VelocityAcquisition Preamble
LINK Establish Required For:Media Access Slot Assignment, Collision, …
NETWORK Communicate Required For:Address Assignment within Address SpaceRouting Informing Other Routers of
Subnet Path
TRANS-PORT
Function Required For:Open Path Negotiate Window SizeACK/NAC Confirm each block
Operations Required Other than Data Transfer
2
4
1
3
12
Connectionless Networks Program Concept
Connectionless Networks Program Concept
•• Reduce Communications Energy Reduce Communications Energy Requirements by at Least 10Requirements by at Least 1022
–– Eliminate Link Acquisition and Eliminate Link Acquisition and Synchronization Synchronization
–– Leverage Platform Awareness to PreLeverage Platform Awareness to Pre--Compute Appropriate Compute Appropriate Communications ParametersCommunications Parameters
–– Integrate ChipIntegrate Chip--Level Power Level Power Management into ProtocolManagement into Protocol
–– Provide AdProvide Ad--Hoc Routing Decisions Hoc Routing Decisions without Routing Info Across Network without Routing Info Across Network in Advancein Advance
–– Provide High Burst Rate, Low Provide High Burst Rate, Low Power/Duty CyclePower/Duty Cycle
Subnet Cluster Routing Node
Connection-Oriented Networks Force Routing Bottlenecks
Ad-Hoc Networks Add Capacity As Node Density Increases
Enabling TransitionEnabling Transition
Acquisition-Less Operation is Key to Achieving True Ad-Hoc Networks
13
CN Key Technical ChallengesCN Key Technical Challenges•• BasebandBaseband Coherent Phase TrackingCoherent Phase Tracking
–– ChallengesChallenges•• Energy Reduction and SpillEnergy Reduction and Spill--Over Due to Phase Tracking Errors Over Due to Phase Tracking Errors •• SelfSelf--Interference Drives Signal Sampling rateInterference Drives Signal Sampling rate
–– ApproachesApproaches•• OversampleOversample BasebandBaseband SignalSignal•• Process Multiple Phase/Epoch Time AssumptionsProcess Multiple Phase/Epoch Time Assumptions•• Use Error Detection Code to Identify “Correct” AssumptionUse Error Detection Code to Identify “Correct” Assumption•• Receiver Powered Only During Receive Epochs, and Processing OnlyReceiver Powered Only During Receive Epochs, and Processing Only After After
Energy DetectedEnergy Detected•• Trades Acquisition/Tracking Transmission Energy for ProcessingTrades Acquisition/Tracking Transmission Energy for Processing
•• Routing ApproachRouting Approach–– Challenge: Develop “Power Aware” MAC and Routing Protocols at ChChallenge: Develop “Power Aware” MAC and Routing Protocols at Chip ip
LevelLevel–– Approach: Many Approaches to MANET Exist Approach: Many Approaches to MANET Exist –– Adapt to Exploit Best Adapt to Exploit Best
Features Features
•• Instantaneous AcquisitionInstantaneous Acquisition–– Challenge: Receivers Must Obtain “Near Perfect” Knowledge of Challenge: Receivers Must Obtain “Near Perfect” Knowledge of
Frequency and Epoch (Block) Time to Initiate AcquisitionFrequency and Epoch (Block) Time to Initiate Acquisition–– Approach : Leverage of State Vector Knowledge and Multiple Approach : Leverage of State Vector Knowledge and Multiple
Hypothesis DetectionHypothesis Detection
14
CN PerformersCN Performers
•• GDDSGDDS
•• HRLHRL
•• RaytheonRaytheon
•• WescommWescomm
•• BAE SystemsBAE Systems
•• BBN TechnologiesBBN Technologies
•• GEGE
15
Disruption Tolerant NetworkingDisruption Tolerant Networking
•• Tactical Networks Do Not Behave Like Office NetworksTactical Networks Do Not Behave Like Office Networks–– Wireless Connectivity Disrupted by Terrain, Weather, Jamming, Wireless Connectivity Disrupted by Terrain, Weather, Jamming,
Access Scheduling, …Access Scheduling, …–– FCS and SUO Both Evidenced Episodic ConnectivityFCS and SUO Both Evidenced Episodic Connectivity
•• Today’s Design Choices Are Not Sufficient to Deal With Today’s Design Choices Are Not Sufficient to Deal With Military NeedsMilitary Needs–– All IP Networks Require:All IP Networks Require:
•• Continuous End to End Path During TransferContinuous End to End Path During Transfer•• Knowledge of Specific Destination Node AddressKnowledge of Specific Destination Node Address•• Routing Information Distributed to Every RouterRouting Information Distributed to Every Router
–– TCP Fails* with Even Short Disconnects and TCP Fails* with Even Short Disconnects and Performs Poorly in Presence of Delay JitterPerforms Poorly in Presence of Delay Jitter
–– UDP Provides no Reliability Services and UDP Provides no Reliability Services and Can not “Hold and Forward”Can not “Hold and Forward”
–– No Internet Protocol can Function End to No Internet Protocol can Function End to End with Tactical ProtocolsEnd with Tactical Protocols
* Stated with Permission of Vint Cerf
Message input
TCP UDP
IP(All Levels)
DTN(overlaying
protocol)
MobileTactical
JTIDS
MIDSJTRS
Interoperable
NetworkCentric
TCP UDP
IP
Toughening the Radio is Not the Answer!Must Toughen Up the Network Itself!
16
DTN Technology ObjectivesDTN Technology Objectives•• IP Routing vs. DTN DeliveryIP Routing vs. DTN Delivery
–– IP (UDP or TCP) Requires IP Continuity IP (UDP or TCP) Requires IP Continuity Between Nodes for Duration of TransferBetween Nodes for Duration of Transfer
–– DTN can operate over nonDTN can operate over non--IP Paths and IP Paths and Hold Data During Periods When All of Hold Data During Periods When All of Path is Not AvailablePath is Not Available
•• DTN Will Manage Impact of Disruption:DTN Will Manage Impact of Disruption:–– Delay ToleranceDelay Tolerance
•• Organize Information flow into BundlesOrganize Information flow into Bundles•• Create “Intelligent Networks” that Can Create “Intelligent Networks” that Can
Manage the Delivery of BundlesManage the Delivery of Bundles•• Allow Messages to Pass through Network Allow Messages to Pass through Network
with Successive Responsibility (Rather with Successive Responsibility (Rather than End to End Acknowledgement)than End to End Acknowledgement)
•• Opportunistically Leverage Connectivity, Opportunistically Leverage Connectivity, Multiple Routes, …Multiple Routes, …
–– Dynamic Networks Naming and RoutingDynamic Networks Naming and Routing•• LateLate--Binding of Bundles (or Packets) to Binding of Bundles (or Packets) to
Specific Nodes or Delivery PathsSpecific Nodes or Delivery Paths•• Avoid Forcing All Parts of the Network to Avoid Forcing All Parts of the Network to
be Aware of All other Partsbe Aware of All other Parts•• Matches Tactical Unit DeploymentMatches Tactical Unit Deployment
17
DTN vs. InternetDTN vs. Internet
All Bundles Include Certificate
NoneTrust
Information-based Bundle
Data-Driven PacketFundamental Unit
Bundle Intermediaries take Delivery Responsibility
None – Applications Must Assure Delivery
Network Delivery Responsibility
Introduces Very Complex Elements
Extremely SimpleNetwork Complexity
May be EpisodicMust be End to EndConnectivity
Operate with Only Local Node/Topology Awareness
Must have Global Awareness of All Nodes/Routes
ScopeDTN EnhancementInternet
18
DTN TechnologiesDTN Technologies
•• BundlingBundling–– Organizing information into “Bundles” that can be Organizing information into “Bundles” that can be
Delivered by NetworkDelivered by Network
•• Fuzzy SchedulingFuzzy Scheduling–– Making Network Routing Decisions in the Presence of Making Network Routing Decisions in the Presence of
Uncertainty about Available or Optimal PathsUncertainty about Available or Optimal Paths
•• Policy Cognitive OperationPolicy Cognitive Operation–– Moving Intelligence into Network to Make “Best” Choices Moving Intelligence into Network to Make “Best” Choices
on Deliveryon Delivery
•• Deferred / Hierarchical Address BindingDeferred / Hierarchical Address Binding–– Enable Network to Deliver Traffic Without End to End Enable Network to Deliver Traffic Without End to End
Address and Routing InfoAddress and Routing Info
19
DTN ParticipantsDTN Participants
•• Jet Propulsion Laboratories (JPL)Jet Propulsion Laboratories (JPL)–– Ongoing Ongoing DelayDelay Tolerant Networking EffortsTolerant Networking Efforts–– SPARTA Providing Security Feature Support to These SPARTA Providing Security Feature Support to These
ProgramsPrograms
•• MITREMITRE–– Ongoing Ongoing DelayDelay Tolerant Networking EffortsTolerant Networking Efforts
•• New DARPA BAA to be ReleasedNew DARPA BAA to be Released–– Industry Day Held 21 JanuaryIndustry Day Held 21 January
MITREMITRE SPARTASPARTA