Developing a Prototype Ground Station for the Processing ...
Transcript of Developing a Prototype Ground Station for the Processing ...
August 11, 2021
Developing a Prototype Ground Station for the Processing,
Exploitation, and Dissemination of pLEO Sensor Data
i
Presented By: Jen Wilbur (SciTec, Inc)
Debi Rose
Dan Rossiter
Paul Wood
Jen Wilbur
David Simenc
Eric Principato
Travis Williams
Jason Hamant
Matthew McHugh
Sander Malmquist
John Maloney
Outline
2
• Introduction
•OPIR Demo
• SAR Demo
• EO/IR Demo
• Conclusions
Introduction
3
• SwRI and teammates SciTec, and Amazon Web Services (AWS)demonstrated a novel, commercial processing, exploitation, and dissemination (PED) prototype to Air Force Space & Missile Systems Center (SMC).
• Objective: Demonstrate a low-latency, horizontally-scalable, PED capability featuring cloud-based processing of data collected by future payloads sensing in multiple modalities hosted on commercial spacecraft and downlinked through commercial gateway injection points. Demo delivery of processed data to tactical users at forward operating locations.
• Addresses a lack of established gateways or processes to ingest data collected from DARPA’s BLACKJACK-capable spacecraft and distribute that data through a commercial gateway to a warfighter in theater.
PED Prototype
PED Overview - Three OPIR Sensors
4
casino2_forecasino2_aft casino1_aft
Receiver
CLIcasino1-aft
AW
S S3
dat
a st
ora
ge –
Encr
ypte
d
Fram
es
CLIcasino2-aft
CLIcasino2-fore
AWS Oregon Region – us-west-2
AWS Ohio Region – us-east-2
Encrypted Sensor Frame Data Files
Data ForwarderDecrypt, virus scan
Raw Sensor Frames with MetaData
AWS GovCloudRegionUS-gov-west-1
Simulated‘High Side’
AWS S3 data storage – Decrypted Frames
Visualization (VEGA)
GovCloud– Encrypted Frames
Dat
a Tr
ansp
ort
Mis
sio
n D
ata
Pro
cess
ing
Frame Processor
Correlation/ Fusion Processors
2d tracklets
3d tracks
Frame Processor
Frame Processor
MDP Ingest/ Balance
Full-frame streaming data
Unified Data Library
Data ForwarderData Forwarder
Assess ability to transport and process full frame, full rate OPIR data from multiple sensors
Warfighter surrogate
Scene Simulation Geometry: OPIR Demo (3 Sensors)
Parameter Value
PRA Terrain DB Korea 300m
PRA Cloud DB Kiev
Local Time 12:00 (noon)
Sensor off-nadir angle
27o
Satellite mean anomaly spacing
9.35o
Sensor size 2048x2048 px
Sim Time 165 s
Frame rate 20 Hz
Frames per sensor 3300 frames3 body-fixed sensors were simulated on 2 spacecraft.
Spacecraft were separated to maximize boresight overlap
5
Simulated Data For the OPIR 2k x 2k Demo
Simulated data provided a wide range of realistic conditions.
Target Signatures
6
Scenes w/ Targets
Transmitter (CLI) – decommutates the CCSDS encoding, initiates the transport process
‘Receiver’ – lambda function - signals forwarder when complete file is available in Receiver region S3
Forwarder – receives files from receiver S3, decrypts, virus scans, places un-encrypted file in Forwarder S3 bucket ready for MDP ingestion
CASINO PED – Data Transport Pipeline
7
Receiver
CLIcasino1-aft
AW
S S3
dat
a st
ora
ge –
Encr
ypte
d
Fram
es
AW
S S3
dat
a st
ora
ge –
Encr
ypte
d
Fram
es
Dat
a Fo
rwar
der
Dat
a Fo
rwar
der
Dat
a Fo
rwar
der
AW
S S3
dat
a st
ora
ge –
Dec
ryp
ted
Fr
ames
Mis
sio
n D
ata
Pro
cess
ing
(MD
P)
Dat
a In
gest
ion
CLIcasino2-aft
CLIcasino2-fore
AWS GovCloud – us-gov-west-1 AWS Oregon Region – us-west-2
AWS Ohio Region – us-east-2
casino2_fore
casino2_aft
casino1_aft
OPIR PED End-to-End Timeline
8
• Average end-to-end timeline for OPIR 2kx2k processing from receipt at gateway to delivery to Mission Data Processing in GovCloud = ~11sec
• End-to-end time includes all data transfer (Xfer), Overhead (OH), and Processing
Test Date CLI Rcv Fwd Total (avg) Min Max
8 Jan 2021 2.46 0.93 1.15 4.54 3.03 16.59
8 Jan 2021 2.66 0.93 2.56 6.20 3.59 34.00
Test Date CLI Rcv Fwd Total (avg) Min Max
8 Jan 2021 2.21 0.92 1.15 4.29 3.29 12.31
8 Jan 2021 2.25 0.93 2.64 5.82 3.65 21.33
OPIR 2kx2k DT Processing Times (seconds)
OPIR 2kx2k Xfer Times (seconds)
OPIR Nominal Test Run Summary
High Level OPIR MDP Architecture
3. BKG Processor – Ingest raw calibrated full-frame imagery,
perform clutter suppression, and output clutter-suppressed full-frame imagery
4. TDE Processor – Ingest clutter-suppressed full-frame
imagery, perform track-before-detect processing, output 2D tracklets
5. CORR Processor – Ingest 2D tracklets from multiple
sensors, perform multi-sensor measurement correlation, output associated measurements
6. FUS Processor – Ingest associated measurements, perform
state vector estimation, output 3D tracks
7. Data analysis/App Dashboard – Interface for
executing and running the CASINO MDP in AWS as well as for analyzing and visualizing results
Full Frame Interface
9
1. Data Ingest – Read CASINO data frames transported to AWS
GovCloud region storage
2. Data Balancer – Route and balance CASINO data to MDP
elements
1
3
4
5
67
2
As Demonstrated End-to-End MDP Timing (2k x 2k)
10
MDP: Elastic Scaling to 30 Sensors (10X Demo)
11
Single Sensor SAR Demo Overview
12
MDP Processing
Encrypted Sensor Frame Data Files
Data Forwarder
MDP Ingest/ Balance
Decrypt, virus scan
Raw Sensor Frames with MetaData (Simulated binary data format for SAR)
‘Low Side’
CASINO simulated binary data →(encrypted → CCSDS packetized → TCP/IP)
CASINO simulated binary data (sensor 2 → (encrypted )
AWS GovCloudRegion
US-gov-west-1
Simulated‘High Side’
AW
S S3
dat
a st
ora
ge –
Encr
ypte
d
Fram
es
GovCloud– Encrypted Frames
AWS S3 data storage – Decrypted Frames
Receiver
POP – IP routing to AWS Cloud Resources
AWSUS-West-2
Oregon(CONUS)
Data Prep/ CLI
Starlink –‘User Terminal’
Local POP
SpaceX Equip Rack
Starlink –‘Gateway’
Redmond WAor Los Angeles CA
Area
Ka Downlink– >10 Gbps
Downlink of Sensor Data from Satellite
CASINO simulated binary data →(encrypted → CCSDS packetized → TCP/IP)
Limited Uplink Bandwidth baseline 10 Mbps
Internet Connection
Commercial Cloud Data Center(Oregon)
Dat
a Tr
ansp
ort
Mis
sio
n D
ata
Pro
cess
ing
Original Purpose/Goals:• Assess ability to transport and process
SAR Data – including transport across Starlink connection
SAR Mission Data Processing Overview
13
SAR MDP Architecture
14
CASINO PED Overview: Single Sensor EO/IR
15
Encrypted Sensor Frame Data Files
Data Forwarder Decrypt, virus scanRaw Sensor Frames with MetaData (Simulated binary data – EO/IR format)
AWS Direct Connect
(100 Gbps)
‘Low Side’
Ground Antenna Site/ Local AWS Data Center(Bahrain on OCONUS)
AWSGround Station
AWS –ME-south-1(OCONUS)
CASINO simulated binary data –EO/IR format → (encrypted →TCP/IP)
CASINO simulated binary data – EO/IR format (sensor 2 →(encrypted )
AWS GovCloudRegionUS-gov-west-1
Simulated‘High Side’
Data Prep / DemoCLI
AW
S S3
dat
a st
ora
ge –
Encr
ypte
d
Fram
es
AWS S3 data storage – Decrypted Frames
Receiver
AWSUS-West-2(CONUS)
Visualization (VEGA)
EO/IR Ingest and Balance
Automatic Target Recognition (ATR)
CCSDS Decomm occurs prior to transfer to Receiver
Commercial Cloud Data Center(Oregon)
GovCloud– Encrypted Frames
Dat
a Tr
ansp
ort
Mis
sio
n D
ata
Pro
cess
ing
Original Purpose/Goals:• Assess ability to transport and process
EO/IR data
EO/IR Demonstration MDP Summary
16
Conclusions
17
• The SwRI Team demonstrated a robust, low-latency solution capable of transporting, processing, and delivering data received from a LEO satellite to a secure cloud processing center, where MDP algorithms were applied to produce actionable information products for the Warfighter• As expected, transfer times are dependent on distance from destination and speed of the
network• We successfully demonstrated MDP of OPIR data that keeps up with the full-frame data rates
expected for “next generation” sensors (low latency), can be scaled elastically to accommodate data streams from multiple sensors (30 demonstrated), and that is modular – supporting multiple missions
• Our PED prototype successfully transported and processed multi-mission data collected by a variety of pLEO sensor types to output information relevant to the Warfighter• OPIR – output 3D tracks in Tactical message formats• SAR – processed I/Q data to generate intuitive, information dense, images• EO/IR – Ingested high spatial resolution imagery and processed with an AI neural net-based ATR
algorithm with cloud masking to enable robust detection and classification of objects of interest