Multidisciplinary System Design Optimization for a Distributed Solar Observation Constellation! Ben Corbin and Ted Steiner 1 October 14, 2014

Transcription

1 Multidisciplinary System Design Optimization for a Distributed Solar Observation Constellation! Ben Corbin and Ted Steiner! October 14 th, 2014! Ben Corbin and Ted Steiner 1

2 Outline! Introduction and Motivation! Mission Concept! Problem Formulation! Modeling and Simulation! Optimization! Conclusions and Future Work!! Ben Corbin and Ted Steiner 2

3 Key Research Motivations! Emergent Capabilities of Distributed Satellites! What scientific goals can be achieved and what scientific knowledge gaps can be filled ONLY through the use of multiple satellites working together?! Emergent Properties of Federated Satellites! How can federated satellites enable or significantly reduce the cost of future science missions?!! What analytical techniques and metrics should be applied in the concept developed phase to properly communicate the value of federated satellite systems?! Ben Corbin and Ted Steiner 3

4 Deep Space Network Congestion! The Deep Space Network (DSN)! Network of 70m and 35m antennas around the world! Network is outdated, expensive to use, and is over-subscribed, but it s the only way to retrieve data from faraway spacecraft! How can we avoid clogging it even more with interplanetary small satellites?! Ben Corbin and Ted Steiner 4

5 ConOps for Interplanetary Network! Small satellites are too small and too far away to communicate with Earth regularly! Relay satellite comes close enough to download data and then forward it to Earth (Store and Forward)! Ben Corbin and Ted Steiner 5

6 Mission Concept! Heliocentric-Orbiting Baseline-Optimized Cosmos Observation Paradigm Ben Corbin and Ted Steiner 6

7 Introduction: HOBOCOP! Heliocentric-Orbiting Baseline-Optimized Cosmos Observation Paradigm Mission A: Solar Observation Campaign! Mission B: Interplanetary data-collection network! Hobo: Small, autonomous satellite collecting time insensitive data! Relay: retrieve data from Hobos and relay it to Earth!!! Ben Corbin and Ted Steiner 7

8 Key Stakeholders and Goals! Solar Physics and Magnetosphere Scientists! 1. Determine the structure and dynamics of the magnetic fields at the sources of solar wind.! 2. Trace the flow of energy that heats the corona and accelerates the solar wind.! 3. Explore dusty plasma near the Sun and its influence on solar wind and energetic particle formation.! 4. Determine what mechanisms accelerate and transport energetic particles.! Federated Satellite Communication Company! 1. Construct a flexible network of satellites to create an interplanetary communications network.! 2. Minimize the cost (mass) of the system while satisfying engineering constraints.! 3. Maximize the total possible data throughput of the system (or other communications metrics to be determined)! 4. Enable future science missions/additional revenuegenerating customers.! Ben Corbin and Ted Steiner 8

9 Architectural Problem Formulation! Problem Statement: To minimize total system mass by optimizing orbit geometry design variables using MSDO techniques while satisfying design constraints.! Objective:! Minimize J(x,p)= {total system mass} Constraints: g(x,p)= { powerused powergen@tem Design Variables:! x= { NRO@NPO@Ecc } h(x,p)= {heatin heatout}=0 Description! Variable! Type! Bounds! Number of Relay Orbits! NRO! discrete! [1, 2, 3, 4, 5, 6]! Relays per Orbit! NPO! discrete! [1, 2, 3, 4, 5, 6]! Relay Orbit Eccentricity! Ecc! continuous! 0.1 to 0.9! Selected Parameters:! p= { Number of Hobos@Data Generation@Minimum Link Rate }={ 72@10 bps@1 Ben Corbin and Ted Steiner 9

10 Example Architectures! NRO = 1! NPO = 2! Ecc = 0.5! NRO = 3! NPO = 2! Ecc = 0.7! NRO = 4! NPO = 4! Ecc = 0.3! Ben Corbin and Ted Steiner 10

11 System Model! Compute minimum total system mass in heliocentric orbit! Ben Corbin and Ted Steiner 11

12 Rearranged System Model! Compute minimum total system mass in heliocentric orbit! Ben Corbin and Ted Steiner 12

13 Model Reduction! Module Arrangement!! Lookup table for orbit parameters! Design & Parameters! Optimizer! Objectives! Lookup distance S1 given x!! Consolidate subsystem optimization! Orbit Module! Comm Module! Subsystem Module! Substitute S1 Ben Corbin and Ted Steiner 13

14 Simulation Models! Orbit Propagation Module! Propagates orbits and calculates distances between satellites!! Communication Module! Minimize: Sum of transmitter and antenna masses! Subject to: Link rate & margin constraints, orbit & comm. parameters!!!! Mass Module! P T ( π D T /λ ) 2 ( λ/4πs1 ) 2 ( π D R /λ ) 2 /k T sys R ( E B / N 0 Minimize: Subsystem masses! Subject to: Power and thermal constraints, design parameters! Ben Corbin and Ted Steiner 14

15 Distance From Relays! S1 Ben Corbin and Ted Steiner 15

16 Max-Min Distance (S1)! Peak at e=0.6 caused by 2:1 resonance Peak at e=0.3 caused by 3:2 resonance Ben Corbin and Ted Steiner 16

17 Design of Experiments! 0.15! Factor 1! Factor 2! Factor 3! Relay Orbits! Relays Per Orbit! Eccentricity! Level 1! 2! 1! 0.20! Level 2! 3! 2! 0.45! Level 3! 4! 3! 0.70! Main Effects on S1! 100 Relay Orbits! Relays Per Orbit! Eccentricity! Main Effects on Total Mass! Effect Level Effect Level Results! Trade between comm. and thermal rejection masses! S1 is chaotic with eccentricity! Resonances disrupt trends! Ben Corbin and Ted Steiner 17

18 Single Objective Optimization (2)! Full Factorial Solution 2,916 Total Designs Global Minimum! NRO = 2, NPO = 1, e = 0.48! Simulated Annealing: Terminated at a design within 1 HOBO Mass = 2.62 kg! kg of the global optimum after 3,496 evaluations.! Relay Mass = 29.2 kg!! Total Mass = 247 kg! Nelder-Mead Simplex: Found the global optimum 5 times out of 25 trial runs, averaging 18.7 iterations each.! Ben Corbin and Ted Steiner 18

19 Alternatives to Optimal Design! NRO = 2 NPO = 1 e = 0.48 NRO = 2 NPO = 2 e = 0.48 Total Mass: 247 kg Total Mass: 304 kg Ben Corbin and Ted Steiner 19

20 Alternatives to Optimal Design! NRO = 2 NPO = 1 e = 0.48 NRO = 3 NPO = 1 e = 0.55 Total Mass: 247 kg Total Mass: 263 kg Ben Corbin and Ted Steiner 20

21 Alternatives to Optimal Design! NRO = 2 NPO = 1 e = 0.48 NRO = 5 NPO = 1 e = 0.48 Total Mass: 247 kg Total Mass: 294 kg Ben Corbin and Ted Steiner 21

22 Multi-Objective Formulation! Problem Statement: To (1) minimize total system mass and (2) maximize total system data throughput by optimizing orbit geometry design variables using MSDO techniques while satisfying design constraints.! Objective:! Constraints: Minimize J(x,p)= { system mass@ data throughput } g(x,p)= { powerused powergen@tem Design Variables:! h(x,p)= {heatin heatout}=0 x= { NRO@NPO@Ecc } Description! Variable! Type! Bounds! Number of Relay Orbits! NRO! discrete! [1, 2, 3, 4, 5, 6]! Relays per Orbit! NPO! discrete! [1, 2, 3, 4, 5, 6]! Relay Orbit Eccentricity! Ecc! continuous! 0.1 to 0.9! Selected Parameters:! Internal Design Variables:! p= { Number of Hobos@Data Generation@Minimum d= { HOBO Link Transmitter Rate }={ 72@10 Power@HOBO bps@1 Ben Corbin and Ted Steiner 22

23 Single-Architecture Pareto Front! Used NBI on a single-architecture to find the true Pareto front! Used surrogate objective to approximate total data throughput! Pareto front is concave due to quadratic communications constraints! NRO = 2! NPO = 1! e = 0.48! Ben Corbin and Ted Steiner 23

24 Multi-Architecture Pareto Front! Combines 2,916 architectures and 61,236 singlearchitecture Pareto points into one tradespace! Lower NRO good for low mass! Higher NRO good for high data rates! Ben Corbin and Ted Steiner 24

25 Optimization vs. Brute Force! Optimization used surrogate objective! Maximize data rate at S1! Brute force used actual objective! Maximize possible data throughput! 3 hours, 61,236 NBI points (complete) vs. 20 hours, 30,000 Monte Carlo points (nowhere near complete)! Ben Corbin and Ted Steiner 25

26 Conclusions! HOBOCOP! Final architecture selection is dependent on stakeholder inputs and needs! Future work will build higher-fidelity models, continue the tradespace exploration, and define more objectives based on system lifecycle properties ( ilities )! MSDO! We don t like discrete variables, long propagation times, or chaotic relationships! Still had success with a variety of optimization methods! Project did not do well with scaling analysis but was excellent for model order reduction, tradespace exploration, and multi-objective analysis! Ben Corbin and Ted Steiner 26

27 Future Work! Science Objectives! Map instrument choices and models to satisfaction of attributes of each science goal! Combine attribute satisfactions and identify Pareto elements! Communications Objectives! Identify and model additional communication system goals and attributes! Consider flexible options and how they can add value over the course of the system lifecycle! Methodology Objectives! Carry out full Responsive Systems Comparison (RSC) method, the marriage of Multi-Attribute Tradespace Exploration (MATE) and Epoch-Era Analysis (EEA)! Identify and quantify instrument synergies and how they affect the value of the scientific return! Ben Corbin and Ted Steiner 27

28 Questions! Ben Corbin and Ted Steiner 28

29 APPENDIX! Ben Corbin and Ted Steiner 29

30 Single-Objective Optimization (1)! Ben Corbin and Ted Steiner 30

31 HOBO Satellites! Similar to a 3μ Cubesat (~3kg)! Primary Instrument: Magnetometer! Assumed: 0.5 kg, 5 W! Actual: kg, 1 W! Small parabolic antenna! Reaction control wheels! Star Tracker! Ben Corbin and Ted Steiner 31

32 COP Satellite! Flying antenna with minimum other subsystems to complete mission! Needs to survive radiation and thermal environment associated with close passes near the Sun! Tradeoffs between high eccentricity/better network performance and more mass due to fuel and thermal and radiation protection! Ben Corbin and Ted Steiner 32

33 Network Simulation NRO = 5, NPO = 3! Ben Corbin and Ted Steiner 33

34 Next Steps in Modeling! Need to develop a communications system performance metric! Possible idea: Total amount of data that can be pushed through the network! Possible requirement: total hard drive space must have asymptotic slope = 0! Build multi-objective tradespace! Ben Corbin and Ted Steiner 34

35 Total Data Metric! Possible data transfer that could be achieved from the HOBOs! Ben Corbin and Ted Steiner 35

36 Hard Drive Example! Data remaining on hard drives increases if link budget is too low or if data generation rate is too high! Example: Data Generated = 1x10 6 bits per day Ben Corbin and Ted Steiner 36

37 Hard Drive Example! Data remaining on hard drives increases if link budget is too low or if data generation rate is too high! Example: Data Generated = 2x10 1x10 6 bits per day Ben Corbin and Ted Steiner 37

38 Future Work Additional Research Questions! How does laser communication change the mass and system performance compared to radio communication?! What advantages do flexible relay orbits have when considering other science mission outside the baseline case?! Examples: NEO exploration, Mars relays! Ben Corbin and Ted Steiner 38

39 Basic Definition! Phase 1: Spacecraft Systems and Performance Models Model the expected performance and other characteristics of a system given a design vector input. Phase 2: Multi-Attribute Utility Theory Determine how stakeholders attain value from a system and be able to rank design alternatives Phase 3: Populate and Explore Tradespaces Model many design vectors and display them on a tradespace, showing comparisons among many architectures Phase 4: Epoch-Era Analysis Consider alternative scenarios and how the systems continue to deliver value in changing exogenous contexts Ben Corbin and Ted Steiner 39

40 Basic Definition! Phase 1: Spacecraft Systems and Performance Models Phase 2: Multi-Attribute Utility Theory Multi-Attribute Tradespace Exploration (MATE) Phase 3: Populate and Explore Tradespaces Responsive Systems Comparison (RSC) Phase 4: Epoch-Era Analysis Ben Corbin and Ted Steiner 40

41 Inputs Basic Overview! Models Outputs This is the new but difficult part. It requires excellent communication with stakeholders and careful understanding of their needs. Stakeholder Attribute Definitions Stakeholder Value Model Multi-Attribute Utility Context Variables System Performance Model Attribute Values Design Variables Intermediate Variables Cost Engineers are REALLY GOOD at doing these things. This is the easy but tedious part Input Model Calculated Parameter Final Output Ben Corbin and Ted Steiner 41

42 RSC and Science Goals! What are the primary science goals? Science Goal #1 What determines how well the goal is met? Attribute #1 What measureable levels give the least and most utility Context 1 min max Science Goal #2 Attribute #2 Context 2 min max Science Goal #3 Attribute #3 Context 3 min max and what contextual changes would your mind about those levels, and how would they change?* Ben Corbin and Ted Steiner 42

43 RSC and Science Goals! Each attribute has a weight Context Science Goal #1 Attribute #1 k = Science Goal #2 Attribute #2 k = Science Goal #3 Attribute #3 k = that may change in different contexts Ben Corbin and Ted Steiner 43

44 RSC and Science Goals! Each science goal has a weight Science Goal #1 k = Science Goal #2 k = Science Goal #3 k = that may change in different contexts Ben Corbin and Ted Steiner 44