seari.mit.edu
SEAri Short Course Series Course: PI.27s Value-driven Tradespace Exploration for System Design Lecture: Lecture 7: Case Studies in Tradespace Exploration Author: Adam Ross and Donna Rhodes Lecture Number: SC-2009-1-7-1 Revision Date: August 11, 2009 A slightly earlier version of this course was taught at PI.27s as a part of the MIT Professional Education Short Programs in June 2009 in Cambridge, MA. The lectures are provided to satisfy demand for learning more about Multi-Attribute Tradespace Exploration, Epoch-Era Analysis, and related SEAri-generated methods. The course is intended for self-study only. The materials are provided without instructor support, exercises or “course notebook” contents. Do not separate this cover sheet from the accompanying lecture pages. The copyright of the short course is retained by the Massachusetts Institute of Technology. Reproduction, reuse, and distribution of the course materials are not permitted without permission.
PI.27s VALUE-DRIVEN TRADESPACE EXPLORATIONFOR SYSTEM DESIGN
Lecture 7Case Studies in Tradespace Exploration
Dr. Donna Rhodes and Dr. Adam M. RossMIT
seari.mit.edu © 2009 Massachusetts Institute of Technology 2
Outline• Case studies• Look out for Artifacts
– Discretization– Modeling– Mathematical
• Find Root Causes of Real Patterns– Physics– Design Differentiators
• Explore the Pareto Front– What is there and Why?
• Explore Alternate Views– A cautionary tale
• Drill down on some designs– Additional insights from more detailed models
seari.mit.edu © 2009 Massachusetts Institute of Technology 3
Example TSE Benefits
• Forces alignment of solutions to needs• Reveals structure of design-value spaces not apparent with few point
designs– Akin to graphing calculator showing function shapes, tradespace gives
insight/intuition into complex design-value space relationships• Facilitates cross-domain socio-technical conversation• Ability to discover compromise solutions
– Beyond “optimized” per stakeholder solutions– Experts often unable to find “suboptimal” solution that may be better
compromise across stakeholders• Structured means for considering large array of possible futures for
discovering robust systems and strategies
The following strengths of TSE were identified by a user of the method
MATE highlights and helps to focus attention on important trades, possibly overlooked by traditional methods
seari.mit.edu © 2009 Massachusetts Institute of Technology 4
Case StudiesName Num DV Num Att Size TradespaceA-TOS 7 2 1380B-TOS 6 5 4033C-TOS NA 5 1X-TOS 9 5 50488SDB 5 2-4 8700+8700+35000SBR 5 7 1872Space Tug 3 3 137TPF 8-10 4-6, 2-3 10611JDAM 8-11 5-8, 1-3 7151RDSS 5,7 10 2340+8640SRS 10 5,5,5 23328x245ORS 6 6 1100+1100+1100HSR 15 20 TBD
There have been 13 case studies as of June 2009
seari.mit.edu © 2009 Massachusetts Institute of Technology 5
A-TOSSwarm making in-situ ionosphere measurements
In Situ Ionospheric Measurements
DESIGN VARIABLES (7)• Bulk Orbit Variables
– Swarm inclination (deg) 63.4– Swarm perigee altitude (km) 200 – 800– Swarm apogee altitude (km) 200 – 800– Swarm argument of perigee (deg) 0– Number of orbit planes 1– Swarms per plane 1
• Swarm Orbit Variables– Subsats per swarm 1 – 26 – Number of subplanes in each swarm 1 – 2 – Number of suborbits in each subplane 1 – 4 – Yaw angle of subplanes (deg: vector) 60– Maximum satellite separation (km) 0.001 – 200
• Non-orbit Variables– Mothership (yes/no)
ATTRIBUTES (2)– “Value” in Low latitude mission, “Value” in High latitude mission
Number of Designs Explored: 1380
0 2 4 6 8 10 120
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Life Cycle Cost, $100M
Tota
l Util
ity
Life Cycle Cost vs. Total Utility (N=1380)
0 2 4 6 8 10 120
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 2 4 6 8 10 120
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Life Cycle Cost, $100M
Tota
l Util
ity
Life Cycle Cost vs. Total Utility (N=1380)
Valu
e
Value
seari.mit.edu © 2009 Massachusetts Institute of Technology 6
B-TOSHills swarm making ionosphere soundings
Top-side sounder swarm
Number of Designs Explored: 4033
0.98
0.985
0.99
0.995
1
100 1000
Lifecycle Cost ($M)
Util
ity
A
EDC
B
DESIGN VARIABLES (6)• Large Scale Arch
– Circular orbit altitude (km) 1100, 1300– Number of Planes 1, 2, 3, 4, 5
• Swarm Arch– Number of Swarms/Plane 1, 2, 3, 4, 5– Number of Satellites/Swarm 4, 7, 10, 13– Radius of Swarm (km) 0.18, 1.5, 8.75, 50
• Vehicle Arch– 5 Configuration Studies Trades payload,
communication, and processing capability
ATTRIBUTES (5)– Spatial Resolution (deg2), Revisit Time (min), Latency (min), Accuracy (deg), Inst.
Global Coverage (%),
seari.mit.edu © 2009 Massachusetts Institute of Technology 7
C-TOSICE design of B-TOS like vehicles*
Number of Designs Explored: 1All dimensions in meters
All dimensions in meters
System Engineer
SubsystemEngineer
SubsystemEngineer
SubsystemEngineer
SubsystemEngineer
SubsystemEngineer
SubsystemEngineer
SubsystemEngineer
ICEMakerPerformance Specifications
Performance EvaluationParameters
Interface Parameters (I.P.s)
I.P.sI.P.s
I.P.s
I.P.s
I.P.sI.P.s
I.P.s
Daughters
MotherDaughters
Velocity
Vector
*Design similar to “D” from B-TOS
seari.mit.edu © 2009 Massachusetts Institute of Technology 8
X-TOSSingle-Vehicle in-situ density measurements
Number of Designs Explored: 50488
DESIGN VARIABLES (9)• Mission Design
– Scenario 1 sat, single launch2 sats, sequential launch2 sats, parallel launch
• Orbital Parameters– Apogee altitude (km) 200-2000– Perigee altitude (km) 150-350– Orbit inclination (deg) 0, 30, 60, 90
• Physical Spacecraft Parameters– Antenna gain low, high– Communication architecture TDRSS, AFSCN– Propulsion type electric, chemical– Power type solar, fuel cell– Delta_v (m/s) 200-1000
ATTRIBUTES (5)– Data lifespan (mos), Sample altitude (km), Latitude diversity (deg), Equatorial time
(hr/day), Latency (hrs)
Total Lifecycle Cost ($M2002)
seari.mit.edu © 2009 Massachusetts Institute of Technology 9
Space TugGeneral purpose orbital service vehicle
Number of Designs Explored: 137
DESIGN VARIABLES (3)• Physical Spacecraft Parameters
– Tugging Capability low, med, high, extreme– Propulsion Type storable-bi, cryogenic, electric, nuclear– Propellant Mass (kg) 0-50000
ATTRIBUTES (3)– Capability (kg), DeltaV (m/s), Fast (y/n)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 500 1000 1500 2000 2500 3000 3500 4000
Cost ($M)
Util
ity (d
imen
sion
less
)
Biprop CryoElectricNuclear
0
10
20
30
40
50
60
70
80
Num
ber O
f Sat
ellit
es
<40
40-4
242
-44
44-4
646
-48
48-5
050
-52
52-5
454
-56
56-5
858
-60
60-6
262
-64
64-6
666
-68
68-7
070
-72
72-7
474
-76
76-7
878
-80
80-8
282
-84
84-8
686
-88
88-9
090
-92
92-9
494
-96
96-9
898
-100
100-
102
102-
104
104-
106
106-
108
108-
110
110-
112
112-
114
114-
116
116-
118 200-300
700-800
1200-1300
Inclination[deg]
70-8060-7050-6040-5030-4020-3010-200-10
0
10
20
30
40
50
60
70
80
Num
ber O
f Sat
ellit
es
<40
40-4
242
-44
44-4
646
-48
48-5
050
-52
52-5
454
-56
56-5
858
-60
60-6
262
-64
64-6
666
-68
68-7
070
-72
72-7
474
-76
76-7
878
-80
80-8
282
-84
84-8
686
-88
88-9
090
-92
92-9
494
-96
96-9
898
-100
100-
102
102-
104
104-
106
106-
108
108-
110
110-
112
112-
114
114-
116
116-
118 200-300
700-800
1200-1300
Inclination[deg]
70-8060-7050-6040-5030-4020-3010-200-10
LEO Targets
seari.mit.edu © 2009 Massachusetts Institute of Technology 10
Terrestrial Planet Finder (TPF) A “Big Science” Project
Number of Designs Explored: 10611
DESIGN VARIABLES (8+2)• Orbit Type {L2, LO, DA}• Num Apertures 4,6,8,10• Wavelength 7,10,20 microns• Interferometer Type {sci, ssi, tsi}• Aperture type {circular optics, strip optics}• Aspect Ratio {multi, const}• Aperture Size Variable• Interferometer Baseline Variable• [Schedule] [0-1 for each obs type]• [Design Lifetime] 5 or 10
ATTRIBUTES (4+2, 2+1)• Science
– Num of Surveys (num), Num Medium Spectroscopies (num), Num Deep Spectroscopies (num), Number of Images (um) [Num Long Baseline Images (num), Num Short Baseline Images (num),
• Agency– Lifecycle Cost ($M), Operational Lifetime (years), [Annual Ops Cost $M)]
From Ross, 2006
“Fixed” Schedule
“Optimal” Schedule
seari.mit.edu © 2009 Massachusetts Institute of Technology 11
Broad-Band Communication Architectures for future communication nets
DESIGN VARIABLES (6)• Altitude• Inclination• Satellites per plane• Number of planes• Power• Antenna Area
ATTRIBUTES (1)• Total System Subscriber Time
[Boeing, 2002]
Many designs explored, emphasis placed on 13 Pareto front architectures
seari.mit.edu © 2009 Massachusetts Institute of Technology 12
Operationally Responsive Disaster Surveillance System
A Multi-Concept Responsive System
DESIGN VARIABLES (5+7)• Aircraft DV
– Configuration Flag 1-13 (3x jet, 2x small prop, 2x med UAV, small UAV, existing Cessna, Orion, Predator, ScanEagle, Global Hawk)
– Gross Weight Flag low, medium, high– Number of Assets 1-6– Payload Type visible, infrared– Aperture Size 0.01, 0.02, 0.04, 0.07, 0.08 m
• Spacecraft DV– Deployment Strategy 1-4 (on-orbit, launch-ready, pre-fab parts, classic design)– Altitude 120-1100 km– Inclination 0,23,90,sun-synch– Number of assets 1-5– Payload Type visible, infrared– Excess Delta-V 600-1200 m/s– Ops Lifetime 5-10 yrs
ATTRIBUTES (10): Firefighter/Owner– Acquisition Cost ($M), Price/day ($K/day), Cost/day ($K/day), Responsiveness (hrs),
Time to IOC (days), Max % of AOI (%), Time to Max Coverage (min), Time between AOI (min), Imaging Capability (NIIRS level), Data Latency (min)
Number of Designs Explored: 2340+8640
Data Latency
Decide to “build”
Time to IOC
Request “service”
Responsiveness
AOI_1 in “view”
AOI_1 “target”
Time between AOI
AOI_2 “target”
Max AOI Covered
Time to Max Coverage
End of “service”
“Service” duration Cost/dayAcquisition Cost
Time of first “need” AOI(s) Type of image(s)Mission Variables (disaster-specific)
Imaging Capability
Price/day
Data Latency
Decide to “build”
Time to IOC
Request “service”
Responsiveness
AOI_1 in “view”
AOI_1 “target”
Time between AOI
AOI_2 “target”
Max AOI Covered
Time to Max Coverage
End of “service”
“Service” duration Cost/dayAcquisition Cost
Time of first “need” AOI(s) Type of image(s)Mission Variables (disaster-specific)
Imaging Capability
Price/day
Witch Creek FireOctober 2007
Image * from http://www.boeing.com/companyoffices/gallery/images/scaneagle/dvd-1390-1.html
*
seari.mit.edu © 2009 Massachusetts Institute of Technology 13
Satellite Radar SystemSatellite-based radar multi-mission system of system
Tj
Epoch jU
0
Tj
Epoch jU
0
Epoch jU
0
U
0
Epoch i
TiU
0
Epoch i
Ti UUUU
0
Epoch i
TiU
0
Epoch i
Ti UUUU
0
Epoch i
TiU
0
Epoch i
Ti UUU
Tj
Epoch jU
0
Tj
Epoch jU
0
Epoch jU
0
Tj
Epoch jU
0
Tj
Epoch jU
0
Epoch jU
0
Peak Transmit Power 1.5 10 20 [KW] 9 9 9 3 1 1 9 9 9 0 1 9 9 9 9 96Radar Bandwidth .5 1 2 [GHz] 9 9 3 3 1 1 9 9 9 0 1 3 3 3 3 66Physical Antenna Area 10 40 100 200 [m^2] 9 9 9 3 1 1 9 9 9 1 1 9 9 9 9 97Antenna Type Mechanical vs. AESA 9 9 9 3 3 1 9 9 9 1 1 9 9 9 9 99Satellite Altitude 800 1200 1500 [km] 9 9 3 9 9 3 9 9 9 9 3 1 1 1 1 85Constellation Type 8 Walker IDs 0 0 1 9 9 3 0 0 3 9 3 9 9 9 9 73Comm. Downlink Relay vs. Downlink 0 0 0 0 0 9 0 0 0 0 9 9 9 3 9 48Tactical Downlink Yes vs. No 0 0 0 0 3 9 0 0 0 0 9 9 9 3 9 51Maneuver Package 1x, 2x, 4x 1 1 1 1 1 0 1 1 1 1 0 9 3 3 3 27Constellation Option none, long-lead, spare 0 0 0 0 0 0 0 0 0 0 0 9 9 9 9 36
Bas
elin
e S
ched
ule
Act
ual S
ched
ule
(Era
)
Tota
l Im
pac
t
Tracking Imaging
Min
. D
isce
rnab
le V
eloc
ity
Num
ber
of T
arge
t B
oxes
Targ
et I
D T
ime
Targ
et T
rack
Life
Trac
king
Lat
ency
MissionATTRIBUTES
ScheduleProgrammaticsCost
Definition RangeVariable Name Min
imum
Tar
get
RC
S
DES
IGN
VA
RIA
BLE
S
Bas
elin
e C
ost
Act
ual C
osts
(Er
a)
Imag
ing
Late
ncy
Res
olut
ion
(Pro
xy)
Targ
ets
per
Pass
Fiel
d of
Reg
ard
Rev
isit
Freq
uenc
y
Satellite Radar SystemProgram Manager
comptroller
Nation
SI&E
SRS Enterprise Boundary
Capital(non‐fungible assets)
Capital(non‐fungible assets)
National Security Strategy/PolicyNational Security Strategy/Policy
Resources(fungible assets)
Resources(fungible assets)
RadarProductRadarProduct
DNI
NGAJ2
Military
USD(I)
ExtendedSRS
Enterprise
SRS Context
OMBCongress
Which SRS Architecture?
R&DR&D Comm/GrndComm/Grnd
Infra‐
Struct.
0 1 2 3
x 104
0
20
40
60
80
100
Par
eto
Trac
e
Des ign ID
im age v. c os t
0 1 2 3
x 104
0
20
40
60
80
100
120
Par
eto
Trac
eDes ign ID
trac k v. c os t
0 1 2 3
x 104
0
20
40
60
80
100
120
Par
eto
Trac
e
D es ign ID
c om bined v. c os t
DESIGN VARIABLES (10)• Orbit DV
– Orbit Altitude 800, 1200, 1500 km– Constellation design 8 Walker IDs
• Spacecraft DV– Antenna Size 40, 80, 100 m2
– Peak power 1.5, 10, 20 kW– Radar bandwidth 0.5, 1, 2 GHz– Comm Arch relay, downlink– Tactical Downlink yes/no– Path enablers (Extra sats, fuel)
EPOCH VARIABLES (6)– Technology avail, Comm infrastructure, Target list, AISR avail,
Environment, Mission priorities
ATTRIBUTES (3 “missions”, 15 total)– Tracking, Imaging, Programmatic
Number of Designs Explored: 23328; Number of Epochs Explored: 245
seari.mit.edu © 2009 Massachusetts Institute of Technology 14
Tradespace Exploration
• Begins with verification of tradespace– Remove errors and biases from model– Gain confidence of results– Identify and compensate for artifacts
• Look at patterns– Display variations in…
• Design variables• Attributes• Single attribute utilities
– Identify expected and emergent trades
seari.mit.edu © 2009 Massachusetts Institute of Technology 15
Artifacts
• Some tradespace “features” may be caused by artifacts of:– Discretization: to keep design space finite, some variables to
which the utility is sensitive may be coarsely surveyed, resulting in a grainy or stripy tradespace
– Modeling: to keep model tractable, some phenomena to which the utility is sensitive may be coarsely modeled, resulting in similar effects
– Mathematics: modeling or utility math may create unexpected or strange-seeming results
• Pathological - unconverged iterations, implicit formulae without solutions
• Unexpected but correct behavior of utility functions, normalizations, etc.
This is not a compete list!
seari.mit.edu © 2009 Massachusetts Institute of Technology 16
Discretization Artifacts (very common)
SPACETUG• General
purpose orbit transfer vehicles
Utility-Cost Discriminated by Tugging Capability
Utility (dimensionless)
0
500
1000
1500
2000
2500
3000
3500
4000
0.0 0.2 0.4 0.6 0.8 1.0
Low Capability Medium CapabilityHigh CapablityExtreme Capability
Bands due to discrete levels of capability
Cos
t ($M
)
seari.mit.edu © 2009 Massachusetts Institute of Technology 17
Discretization Artifacts (very common)
SPACETUG• General
purpose orbit transfer vehicles
Utility-Cost Discriminated by Tugging Capability
Utility (dimensionless)
0
500
1000
1500
2000
2500
3000
3500
4000
0.0 0.2 0.4 0.6 0.8 1.0
Low Capability Medium CapabilityHigh CapablityExtreme Capability
Bands due to discrete levels of capability
Cos
t ($M
)
seari.mit.edu © 2009 Massachusetts Institute of Technology 18
Modeling Artifacts: B-TOS Swarm
0.98
0.985
0.99
0.995
1
100 1000
Lifecycle Cost ($M)
Util
ity
A
EDC
B
Max Baseline Length
• Design Vector includes number of Vehicles in Swarm
• Vehicles populate rings in Hills orbit around central vehicle
• Model assumes only FULL rings add to quality of data
Ring
Vehicle
• Tradespace shows steps in utility corresponding to filled rings
• Designs with unfilled rings have less utility at the same cost
Question - is this modeling assumption correct??
seari.mit.edu © 2009 Massachusetts Institute of Technology 19
• B-TOS utilities run from .98 to 1• Intuition (incorrectly) concludes all designs are good• Artifact of Utility Function
– One attribute weighted very highly by user is always satisfied
– Non-linear aggregate utility function creates utilities shown
– Could have removed attribute from tradespace to rescale utility (made a constraint)
0.98
0.985
0.99
0.995
1
100 1000
Lifecycle Cost ($M)
Util
ity
A
EDC
B
1.000
0.995
0.990
0.985
0.980Lifecycle Cost ($M)100 1000
Utility
Mathematical Artifacts
seari.mit.edu © 2009 Massachusetts Institute of Technology 20
Root Causes of Patterns
• Physics– Simple patterns can be seen by looking at variation in one
design variable (e.g. space tug fuel load vs. rocket equation)– Some patterns are more difficult to discern - detective work
(e.g. drag/altitude/life/fuel load trades in X-TOS)– Often confounded with modeling artifacts
• Design discriminators– Obvious good/bad patterns (some design choices
dominated) - be careful of these– Trades (design choices in different parts of the trades space
and/or Pareto Front)• Decision maker needs
– May require impossible/unrealistic performance
This is not a compete list!
seari.mit.edu © 2009 Massachusetts Institute of Technology 21
Understanding Limiting Physical or Mission constraints
Hits a “wall” of either physics (can’t change!) or utility (can)
0.00
500.00
1000.00
1500.00
2000.00
2500.00
3000.00
3500.00
4000.00
0.00 0.20 0.40 0.60 0.80 1.00
Low BipropMedium BipropHigh BipropExtreme BipropLow CryoMedium CryoHigh CryoExtreme CryoLow ElectricMedium ElectricHigh ElectricExtreme ElectricLow NuclearMedium NuclearHigh NuclearExtreme Nuclear
SPACETUG• Lines show increasing
fuel mass fraction• Past a certain point,
add only cost (no more utility)
• Some of this is Rocket Equation (big vehicles need more fuel to push themselves around
• Some is “maxing out” of utility function -extra fuel adds no further utility
Cos
t ($M
)
Utility-Cost Discriminated by Capability, Prop Subsys, and Fuel Mass Fraction
Utility (dimensionless)
Prop SubsysLowMediumHighExtreme
CapabilityBipropCryoElectricNuclear
seari.mit.edu © 2009 Massachusetts Institute of Technology 22
Understanding Trades
Total Lifecycle Cost
($M2002)
X-TOS single vehicle• Pareto designs collect low
altitude data (most valuable) at the expense of lifetime
• On the Pareto Front, a complex trade of altitude, lifetime, fuel load, and cost
• Depends on the details of the utility functions -different trades for different preferences
Each point is a specific design
Pareto Front of “best” designs “Dominated”
designs are inferior in cost or utility
Warning: “good” is not always in the upper left (read the axes!)
“Good” Direction
seari.mit.edu © 2009 Massachusetts Institute of Technology 23
Confounding of Real and Artificial Effects
0.98
0.985
0.99
0.995
1
100 1000
Lifecycle Cost ($M)
Util
ity
A
EDC
B
Max Baseline Length
• Vehicles populate rings in Hills orbit around central vehicle
• Model assumes only FULL rings add to quality of data
• Real physics DOES favor filled rings, but is it this binary?
Ring
Vehicle
• Tradespace shows steps in utility corresponding to filled rings
• Designs with unfilled rings have less utility at the same cost
Question - how much is modeling, how much real physics?
seari.mit.edu © 2009 Massachusetts Institute of Technology 24
Design Choices
Number of Designs Explored: 50488
Total Lifecycle Cost ($M2002)
• Full X-TOS tradespace includes concepts with multiple vehicles
• Multi-vehicle designs add a lot of cost, some utility; might prompt decision-maker reassessment
• Be careful: reassess in terms of attributes, and check DM set, before making “easy” decisions
Single VehicleDesigns
Multi-VehicleDesigns
seari.mit.edu © 2009 Massachusetts Institute of Technology 25
Architecture Tradeoffs
TPF• Terrestrial Planet
Finder - a large astronomy system
• Connected and separated architectures clearly separated into to “islands”
• Separated architectures only appropriate for large, high-performance, expensive mission
Image from Beichman et al, 1999, tradespace from Jilla, 2002
Separated
Connected
seari.mit.edu © 2009 Massachusetts Institute of Technology 26
Tracing the Pareto Front
• Examining the designs on the Pareto Front yields information on:– Best designs in various regions - often a region is
dominated by a set of similar designs– Trades in various regions - within each region, a
dominant cost/utility trade may exist• Look for artifacts, physical limitations, and
design decisions behind the observed patterns
This is not a compete description!
seari.mit.edu © 2009 Massachusetts Institute of Technology 27
Different Parts of the Pareto Front have Different “Best” Concepts
SPACETUG• General
purpose orbit transfer vehicles
seari.mit.edu © 2009 Massachusetts Institute of Technology 28
Tracing the Space Tug Pareto Front
• Tenders:– Small, light, inexpensive vehicles– Low capability– Any propulsion except nuclear– Trade is utility vs. cost, driver is fuel load
• Cruisers– Small, light, inexpensive vehicles– But high Delta-V– Only possible with high Isp propulsion (electric)
• Monsters– Desire for large, capable, fast vehicles can only be met by
using high thrust, high Isp propulsion (nuclear)– Trade is capability vs. cost, driver is manipulator mass
seari.mit.edu © 2009 Massachusetts Institute of Technology 29
TPF Pareto Front
Many different designs on the Pareto Front
Several detailed studies were of dominated solutions
seari.mit.edu © 2009 Massachusetts Institute of Technology 30
Terrestrial Planet Finder Pareto Designs
# “Images” LCC ($B) Orbit (AU) # Apert.’s Architecture Apert. Diam. (m)
502 0.743 1.5 4 SCI-1D 1 577 0.762 2.0 4 SCI-1D 1 651 0.767 2.5 4 SCI-1D 1 1005 0.768 1.5 4 SCI-1D 2 1114 0.788 2.0 4 SCI-1D 2 1171 0.790 2.5 4 SCI-1D 2 1195 0.807 1.5 6 SCI-1D 2 1292 0.811 1.5 6 SCI-2D 2 1317 0.830 1.5 8 SCI-1D 2 1424 0.836 2.0 4 SCI-1D 3 1426 0.838 1.5 8 SCI-2D 2 1464 0.867 2.5 6 SCI-2D 2 1631 0.877 1.5 6 SCI-1D 3 1684 0.881 1.5 6 SCI-2D 3 1687 0.932 2.0 6 SCI-1D 3 1828 0.936 2.0 6 SCI-2D 3 1881 0.980 1.5 8 SCI-2D 3 1978 0.982 1.5 6 SCI-1D 4 2035 1.086 2.0 8 SCI-2D 3 2132 1.112 1.5 8 SCI-1D 4 2285 1.120 1.5 8 SCI-2D 4 2328 1.190 2.5 6 SCI-2D 4 2398 1197 3.0 6 SCI-2D 4 2433 1.212 4.0 6 SCI-2D 4 2472 1.221 4.5 6 SCI-2D 4 2482 1.227 5.0 6 SCI-2D 4 2487 1.232 5.5 6 SCI-2D 4 2634 1.273 2.5 8 SCI-2D 4 2700 1.280 3.0 8 SCI-2D 4 2739 1.288 3.5 8 SCI-2D 4 2759 1.296 4.0 8 SCI-2D 4 2772 1.305 4.5 8 SCI-2D 4 2779 1.312 5.0 8 SCI-2D 4 2783 1.317 5.5 8 SCI-2D 4 2788 1.569 3.0 6 SSI-2D 4 2844 1.609 3.5 6 SSI-2D 4 2872 1.655 4.0 6 SSI-2D 4 2988 1.691 2.0 8 SSI-1D 4 3177 1.698 2.5 8 SSI-1D 4 3289 1.739 3.0 8 SSI-1D 4 3360 1790 3.5 8 SSI-1D 4 3395 1.850 4.0 8 SSI-1D 4 3551 1.868 2.5 10 SSI-1D 4 3690 1.919 3.0 10 SSI-1D 4
Family4 ap.
SCI-1D1 m Diam.
Family8 ap.
SCI-2D4 m Diam.
Family8 ap.
SSI-1D4 m Diam.
Intersection of MultipleFamilies
Family4 ap.
SCI-1D2 m Diam.
Family6 ap.
SCI-2D4 m Diam.
Family6 ap.
SSI-2D4 m Diam.
Family10 ap.SSI-1D
4 m Diam.
Transition fromSCI to SSIDesigns
Mission Cost& Performance
Low
Medium
High
seari.mit.edu © 2009 Massachusetts Institute of Technology 31
Understanding the TPF Pareto Front
• A variety of connected architectures dominate the cheaper portion of the front– Trade is adding reflector numbers and size
• Separate architectures dominate the expensive high-performance systems– Trade is increasing numbers of reflectors
Remember that the Pareto Front is not the only part of the tradespace
we care about
seari.mit.edu © 2009 Massachusetts Institute of Technology 32
Alternate Views
• Often views other than the simple Utility/Cost plots are needed
• There are MANY possible views– Single attribute utilities vs. cost or design vector elements– Single attribute utilities vs. each other– Plots for various stakeholders (differing utility functions)
• Use to explore suspected issues, display issues, conflicts, and trades for decision makers
This is not a compete list!
seari.mit.edu © 2009 Massachusetts Institute of Technology 33
Clashing Stakeholders in A-TOS
Clashing stakeholder needs discovered and illustrated
A-TOS• In-situ
measurement of upper atmosphere
• Two missions-high and low
• Plot shows best designs for low latitude missions ignore high latitude mission (!)
• Fundamental disconnect - these missions should not share a vehicle.
seari.mit.edu © 2009 Massachusetts Institute of Technology 34
Terrestrial Planet Finder (TPF)A cautionary tale…
From Jilla, 2002
Jilla (2002) study used a single attribute (number of images)
This was aggregated from a number of different image types, for different science stakeholders
Study showed viable (if expensive) concepts
A “rich” tradespace used in several studies
Single attribute
seari.mit.edu © 2009 Massachusetts Institute of Technology 35
Terrestrial Planet Finder (TPF) A cautionary tale…
• Recast in terms of 4 separate image types
• NO designs could meet minimal acceptable performance for some image types!
• Examining alternate view (individual attributes) exposed both the problem, and paths in the direction of solutions
This is information that decision makers need up front…
seari.mit.edu © 2009 Massachusetts Institute of Technology 36
Looking at Individual Attributes
From Ross, 2006
• Look at one attribute (image type) at a time
• For short baseline attribute, designs exist within the acceptable range (0<utility<1)
• For long baseline attribute, no design meets minimum requirements
Need to:• Look at a wider design space for
solutions• Re-assess stakeholder needs
Shor
t bas
elin
e im
ages
Long
bas
elin
e im
ages
Lifecycle Cost ($M)
Lifecycle Cost ($M)
“best”
“worst”
“best”
“worst”
seari.mit.edu © 2009 Massachusetts Institute of Technology 37
Expanding the Tradespace
From Ross, 2006
• In original studies, imaging sequence was assumed to be fixed
• Instead, pick optimum sequence for each design
• Move from constants vector to intermediate variable
• Result is improvement in performance (though in this case not enough)
“Fixed” Schedule
“Optimal” Schedule
Shor
t bas
elin
e im
ages
Long
bas
elin
e im
ages
Lifecycle Cost ($M)
Lifecycle Cost ($M)
“best”
“worst”
“best”
“worst”
seari.mit.edu © 2009 Massachusetts Institute of Technology 38
Deeper Exploration of Promising Designs
• More designs in interesting regions• Integrated Concurrent Engineering (ICE) to
flesh out designs• Reality / feasibility check• Details may lead to further insights
– Configuration – Component availability and capabilities– External interfaces (e.g., TDRSS)
• Iterating on ICE designs may increase utility
seari.mit.edu © 2009 Massachusetts Institute of Technology 39
Example: Bipropellant GEO Tug
• Approx. 1300 kg dry mass, 11700 kg wet mass• Quite big (and therefore expensive); not very practical (?);
Scale for all images:
black cylinder is 1 meter long by
1 meter in diameter
ManipulatorSystem
Solar Panels
Spacecraft Busw/subsystems
Propulsion Systemw/fuel
seari.mit.edu © 2009 Massachusetts Institute of Technology 40
Electric Propulsion RT GEO Tug
• Approx. 700 kg dry mass, 1100 kg wet mass• Includes return of tug to safe orbit• A reasonable, versatile system
The “Electric Cruiser”
seari.mit.edu © 2009 Massachusetts Institute of Technology 41
Bi-prop Tender Designs
• Lower Utility, lower cost systems • Can’t go to GEO (though can work there if inserted)• 700-1000 kg dry mass; 1000-4000 kg wet mass• A family of potential vehicles with reasonable sizes and mass
fractions
For Space Tug, ICE confirms general conclusions of MATE study
seari.mit.edu © 2009 Massachusetts Institute of Technology 42
Exploring X-TOS Designs in Detail (with ICE)
• Original X-TOS tradespace included designs with unrealistic form factors - “telephone pole sats”
• Discovered in ICE sessions
• MATE models modified to eliminate them
• Detailed ICE iterations resulted in designs not in original MATE enumeration
• Added to modified MATE tradespace
seari.mit.edu © 2009 Massachusetts Institute of Technology 43
Baseline X-TOS Design
• Est. Cost: $71.7 M• Utility*: 0.705• Wet Mass: 449.6 kg• Dry Mass: 188.9 kg• Lifetime: 0.534 years• Orbit: 185 km circular• LV: Minotaur I
* Denotes “Original” User Utility
This design was not on the original tradespace
00.05
0.10.15
0.20.25
0.30.35
0.40.45
Latency Latitude EquatorTime
Lifespan Altitude
Weight Factors of each Attribute (k values)
Original Revised
seari.mit.edu © 2009 Massachusetts Institute of Technology 44
X-TOS Last Design• Est. Cost: $75.0 M• Utility*: 0.556• Wet Mass: 324.3 kg• Dry Mass: 205.5 kg• Lifetime: 2.204 years• Orbit: 300 km circular• LV: Minotaur I
* Denotes “Revised” User Utility
This design motivated by revised user preferences
00.05
0.10.15
0.20.25
0.30.35
0.40.45
Latency Latitude EquatorTime
Lifespan Altitude
Weight Factors of each Attribute (k values)
Original Revised
seari.mit.edu © 2009 Massachusetts Institute of Technology 45
Good Ideas for the Art of Tradespace Exploration
• Look out for artifacts• Find root causes of real patterns• Explore the Pareto Front• Examine alternate views
– use to find or explain issues and try solutions• Drill down on some designs
Remember goal: turn data generated by model into knowledge for decisions makers.
Back up
seari.mit.edu © 2009 Massachusetts Institute of Technology 47
B-TOS Swarm Geometry• Max baseline length is defined
by desired angle of arrival accuracy (.0005 degrees)
• Minimum baseline length limited by beacon frequency (100 MHz)
• Swarm suborbit spacing is a factor of 5.7 and defined by:• Phase error of the swarm
• Frequency of the beacons
• Filling the baselines ensures no ambiguity in the angle of arrival measurement
Mothership
Daughterships
Swarm Suborbits
Max Baseline Length
seari.mit.edu © 2009 Massachusetts Institute of Technology 48
B-TOS Frontier Architectures
Point A B C D EAltitude (km)Num of PlanesSwarms/Plane 1 1 1 1 2Satellites/Swarm 4 7 10 13 13Swarm Radius (km) 0.18 1.5 8.75 50 50Functionality Study
<-- 1100 --><-- 1 -->
<-- #5 -->
Point A B C D EAltitude (km)Num of PlanesSwarms/Plane 1 1 1 1 2Satellites/Swarm 4 7 10 13 13Swarm Radius (km) 0.18 1.5 8.75 50 50Functionality Study
<-- 1100 --><-- 1 -->
<-- #5 -->
StudyType M DNumber 1 3+Payload (Tx) Yes NoPayload (Rx) Yes YesProcessing Yes NoTDRSS Link Yes NoIntra-Swarm Link Yes Yes
5StudyType M DNumber 1 3+Payload (Tx) Yes NoPayload (Rx) Yes YesProcessing Yes NoTDRSS Link Yes NoIntra-Swarm Link Yes Yes
5
seari.mit.edu C-TOS Technical Summary© 2009 Massachusetts Institute of Technology 49
C-TOS Mission Requirements• Altitude: 1100 km• Orbit inclination: 63.4°• Swarm shape is 2:1 ellipse• Swarm radius: (40° angular spacing between all daughters)
– inner ring (3 spacecraft) full major axis 45 km– outer ring (6 spacecraft) full major axis 90 km
• Lifetime: 18 month reliability requirements– mother; 95% (consumables for 5 years)– swarm
• 90% full swarm (9 daughters and one mother)• 95% (2 inner ring daughters, 5 outer ring, and one mother)
• Data latency: 45 minutes from payload to ground• Not designed to cost, parametric cost models• Payload and interface defined by sponsor• Daughters replaceable
seari.mit.edu C-TOS Technical Summary© 2009 Massachusetts Institute of Technology 50
C-TOS Swarm Geometry Requirements
Daughters
MotherDaughters
• All spacecraft in the swarm are coplanar• The velocity vector is on the major axis• The swarm plane is perpendicular to the surface of the earth (90° roll about the
velocity vector)
Velocity
Vector