Upload
vuque
View
233
Download
2
Embed Size (px)
Citation preview
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
1
Continuous Biopharmaceutical Manufacturing: Modeling, Design,
and Fully Automated Control
Richard D. Braatz
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
2
Background: Biopharmaceuticals Manufacturing
• Products derived from biological organisms for treating or preventing diseases
• Hundreds of approved products on the market
• Over 7000 medicines in development
• Usually any off-spec material requires rejection of entire lot
Deloitte (2016). 2016 Global Life Sciences Outlook: Moving Forward with Cautious Optimism. Deloitte LLP, Boston, MA.Informa (2016). Pharmaprojects Pharma R&D Annual Review 2016. Informa PLC, London, United Kingdom.
USD in billions
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
500
400
300
200
100
0
% non‐bio drug candidates in pipeline
% bio drug candidates in pipeline
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
1009080706050403020100
Current State of Biopharmaceutical Manufacturing
• Sequence of batch unit operations
• Upstream: cell culture and harvest
• Downstream: purification using chromatography, filtration, diafiltration
• Bioreactor has T, dissolved oxygen, and maybe pH controls
• Minimal modeling/controls
Kelley, B. (2009). Industrialization of mAb Production Technology: The Bioprocessing Industry at a Crossroad. MAbs, 1, 443-452.Shukla, A. A., Thömmes, J. (2010). Recent Advances in Large-Scale Production of Monoclonal Antibodies and Related Proteins. Trends in Biotechnology, 28, 253-261.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
3
Product manufacturing:Allston, MA USA
Product QC: Haverhill, UK
Fill/finish: Waterford, IE
Cerezyme patients distributed worldwide
Manufacturing Biologic Drugs Today
Motivation: Biopharmaceuticals Manufacturing
• Over 7000 medicines in development
• Industry driven to look for new technology to increase flexibility and reduce costs
continuous flow
novel process designs
Deloitte (2016). 2016 Global Life Sciences Outlook: Moving Forward with Cautious Optimism. Deloitte LLP, Boston, MA.Informa (2016). Pharmaprojects Pharma R&D Annual Review 2016. Informa PLC, London, United Kingdom.
USD in billions
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
500
400
300
200
100
0
% non‐bio drug candidates in pipeline
% bio drug candidates in pipeline
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
1009080706050403020100
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
4
Moving Towards Continuous-flow Operations
http://www.drugdevelopment-technology.com/contractors/contract_research/biovian/biovian1.htmlhttp://www.genengnews.com/gen-articles/novasep-morphs-into-full-service-company/3048/
• Move manufacturing from batch to integrated continuous operations with minimal holdup in between
• Continuous flow enables
higher process flexibility, consistency, and volumetric capacity
lower equipment cost and operational complexity
tighter specifications on product quality
→ need for process modeling/design/controls
Technology Directions
• Increased understanding and optimization of each unit operation and component
• Fully automated microscale technologies for high-speed continuous process development
• Availability of plug-and-play modular unit operations with integrated process control and monitoring systems to facilitate straightforward deployment in the laboratory
• Dynamic mathematical models for unit operations and entire plant to support process development and plant-wide control
• Easy-to-use model-based control technologies for optimizing startup, changeover, and shutdown
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
5
Design of Control Systems Based on “Virtual Plant”
• Constructed from first-principles models wherever possible, grey-box models where necessary
• Highest complexity models used for the invention and optimization of process designs and development
• Lower complexity plant-wide model runs in parallel with fully automated process operations, for process control and quality and equipment condition monitoring
Micromixer
pH
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
6
Towards Biomanufacturing on Demand (BioMOD)
• Enable flexible methodologies for genetic engineering/modification of microbial strains to synthesize multiple and wide-ranging protein-based therapeutics
• Develop flexible & portable device platforms for manufacturing multiple biologics with high purity, efficacy, and potency, at the point-of-care, in short timeframes, when specific needs arise
• Include end-to-end manufacturing chain (including downstream processing) within a small-scale fluidics-based platform
J.C. Love, Towards making biologic drugs on demand, 4th Int. Conf. on Accelerating Biopharmaceutical Development, January 25, 2015
Design Requirements Patient
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forRelease
Not OK
☐
Final Product
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
J.C. Love, Towards making biologic drugs on demand, 4th Int. Conf. on Accelerating Biopharmaceutical Development, January 25, 2015
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
7
Biologic Drugs Produced
Human growth hormone Interferon–α2b
Used for treatment of growth disorders
Used for treatment of cancers and viral infections
Rationale for Pichia pastoris asa Microbial Host for Biosimilar Products
Advantages from a regulatory perspective• Many products on market or in late-stage development
• Reduced risk for viral contamination
• Human-like post-translational modifications (folding, glycosylation, etc.)
Technical benefits• Genetically stable organism
• High-density cultivation (>70% biomass)
• High yields of secreted proteins (up to ~15 g/L)
• Limited host cell protein (HCP) profile (eases burden on downstream)
• Amenable to freeze-drying
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
8
Plant-wide Control Approach
• Build first-principles dynamic modelsfor each unit operation (UO)
• Design control system for each UOto meet “local” material attributes
• Evaluate performance in simulationsand propose design modificationsas needed
• Implement and verify the control system for each UO
• Design and verify fully automated plantwide control system to ensure that the product quality specifications are met
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Plant-wide Control Approach
Characteristics of InSCyT• Multi-product manufacturing plant• Continuous & discrete operations• All of the process characteristics
articulated by Lee/Weekman 1976, Foss 1983, and more
• No SS & must align with regulatory requirements (no off-spec product)
Approach adapted from the chemical industry• Employing systematic & modular design of plantwide control strategies
for large-scale manufacturing facilities (Stephanopoulos/Ng, JPC 2000)• Using algorithms that can handle nonlinearities, uncertainties,
distributed states, time delays, unstable zero dynamics, constraints, and mixed continuous-discrete operations
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
9
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forRelease
Not OK
☐
Final Product
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
Computational fluid dynamics (CFD) simulations for greater insight into bioreactor operation
CAD drawingSteady‐stateflow profiles
Bioreactor
• Solved for steady-state flow profiles within the reactor to understand mixing and species heterogenity
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
10
Bubble Size Distribution Predicted by Multiphase Simulation
Plot shows average size of bubbles, red being the largest
Simulation identifies regions of high turbulence dissipation for bioreactor design modifications
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
11
Robust Split-range Control of Bioreactor
• Designed hybrid robust controllers that switch during each manipulated variable change
• Reduced variability by optimized switching
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forRelease
Not OK
☐
Final Product
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
12
Chromatography Modeling
Nonlinear PDEs
• Species mass balance in liquid phase
• Adsorbed phase balance
ci
tot
ci
z
1
Pe
2 ci
z2 qi
pore
qi
kads ,iciQ
i kdes,iqicsalt i
*( )ii ik q
Inlet Outlet
SorptionConvection &Dispersion
Interstitial volume Porous bead
Pore diffusion
Film mass transfer
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Chromatography Modeling
Nonlinear PDEs
• No steady-state, always in transient mode
• No measurements in the column, large time delay to exit measurement
• Switch between multiple sets of nonlinear PDEs
• Practice is to run open loop or switch on exit concentrations
Inlet Outlet
SorptionConvection &Dispersion
Interstitial volume Porous bead
Pore diffusion
Film mass transfer
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
13
Hybrid Simulation of Chromatography Operation
• Four-stage bind-and-elute operation mode
# Stage Time Protein Conc. Salt Conc.
1 Load tl cp,i cs,l
2 Wash tw 0 cs,w
3 Purge tp 0 cs,e
4 Elute te 0 cs,e
# Specified by upstream units
# #
→ Sensors at entrance and exit only, i.e., feedback limited→ Open-loop optimal control can be used to determine
“optimal” operation based on first-principles model
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Multiple Performance Objectives
Performance Objective
Description
Purity
Recovery
Exit Concentration
Productivity
0
,0 0
,
,
e
e e
pdt elution
pdt elution cont elution
t
t t
c dt
c dt c dt
,
,
0
et
pdt elution
pdt loal d
c dt
t c
,0
e
pdt elution
t
e
c dt
t
,0
e
pdt elution
l w p e d
tQ c dt
t tt t t
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
14
Region oflow quality
localextrema
Global optimum
Multiobjective Optimization
• Purity, recovery, exit concentration, and productivity are competing objectives
• Look to maximize exit concentration & productivity while constraining purity to be >99% and recovery to be >90%
• Multiobjective problem solved using stochasticoptimization to avoid being trapped in low-quality local extrema
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Pareto Optimality (Tradeoff Curve)
Dilution Factor
Curves are optimal tradeoff for various feed concentrations
Select optimal values in a lookup table from measured feed concentration
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
15
Column Tanks 1 2
Downstream
Affinity Chromatography
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
1 2 1 2
Polishing Membrane #1 Tanks
Polishing Membrane #1
Polishing Membrane #2
Waste Waste Waste
Polishing Membrane #2 Tanks
UF/DF Membrane
Waste
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forRelease
Not OK
☐
Final Product
On-line Reactor Control
Perfusion
Fill
Elute
Real-TimeProcess
Data
Buffer makeup unit to avoid carrying all possible buffers on the platform
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Batch Buffer Makeup “UO” Modeling
pH Model (Waller, Kravaris, ...)
Method of reaction invariants
Mass balances are linear equations
pH is a static nonlinear function of the invariants
Multicomponent Mixing RuleRequires estimation of effective ionic radius from data
Single Component Conductivity First-principles Model
Flowrates Fin are the manipulated variables
Semi-batch Tank Reaction Invariant Species Conservation
dV
dt Fin
dN
dt Finwin ;
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
16
Many Systems and Control Problems Appear in the Buffer “UO”
• Dynamics are highly nonlinear
• Slow mixing and uncertain parameters
• Model-based control can overcome these issues by predicting the effect of inputs on the process
– Need good parameter values based on process data
– Need to develop robust control strategies that operate well under uncertainty
Design of experiments for conductivity
pH control in micromixersexamples
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Example of pH Control in Buffer Mixing “UO”
• Use of in-line micromixers allows for process intensification
• pH modeled using the reaction-invariant method
• Material balances modeled using tanks-in-series
Micromixer
pH
Fw, ww
Fb, wb
Fp, wp
Fs, ws
w1 w2 wi wn
11
1 1
1 tin
Fd
dt V V F
wW w
T
w b p sF F F F F
t w b p sF FF F F 1 )(ti i
i
id F
dt V w ww
[ ]in w b p sW w w w w pH ( )nf w
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
17
Direct PID vs Reaction-invariant PID Control
• Direct PID
• Reaction-invariant PID
Slow response
Marginally stable
Fast response
Slow response
PID based on the reaction invariants improves performance across pH range
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Reaction-invariant Controller:Performance Under pH Model Uncertainty
• No model uncertainty
• 20% error in pKas within pH model
InstabilitySlow response
Caveat: Reaction‐invariant control requires accurate models to achieve good
performance
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
18
Model Adaptation of Reaction-invariant Control via Dual Micromixer Configuration
Micromixer
pH
Micromixer
pH
First micromixer used to excite system to enable accurate parameter estimation of pH model
Second micromixer used to perform corrections to the output of the first micromixer and effect setpoint tracking
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Micromixer #1
Micromixer #2
Dual Micromixer Configuration
Input signal to generate data for model adaptation
First micromixer used to excite system to enable accurate parameter estimation in pH model
Second micromixer used to perform corrections to the output of the first micromixer and effect setpoint tracking
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
19
Closed-loop Performance Comparison
• Single micromixer without model adaptation
• Dual micromixer with model adaptation
Model adaptation allows for rapid response in spite of high uncertainty
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
20
Comments on the State-of-the-Art in Modeling, Control, and Monitoring Technology
• The best commercial plant simulation software handles nonlinearities, time delays, unstable zero dynamics, constraints, mixed continuous-discrete operations, and some uncertainty analysis methods (e.g., Si, MC)
• Distributed states facilitated by various math tricks, e.g., moment analysis, transform methods, characteristics
• More sophisticated uncertainty analysis tools can be wrapped around or integrated into such software
• MPC and process monitoring technology are available
• Currently developing methods to simplify tuning and improve performance of model-based control and monitoring systems
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
21
Summary
• Described some research challenges and opportunities in the manufacturing of biopharmaceuticals
• Discussed control technology within the context of a fully automated biopharmaceuticals-on-demand platform
• Model-based control solutions are a mix of nonlinear feedback & feedforward strategies
• Described feedforward controls and an adaptive control strategy motivated by a specific application need
• Described state-of-the-art in process control technology
Closing Comments
• Small-molecule pharma systems engineering went from very little in the late 1990s to becoming well established
• I believe that applications in biologic manufacturing will mirror the rise in small-molecule pharmaceuticals that occurred over the last 20 years, but will be faster
D2
D1
R2
R1LC LC
M1 M2 S1
C1 C2
W1
LC LC
C3 C4
LC
LC LC
S3
S4W2
M3 M4
M5
S5S6 E1 CS
CAT
A
B
S2
S1
S3
S1
PU1
PU2
PU4
PU3
S1
S1
S1
C
D
E
S1
EX1
EX2FP
LC
S1
FCsp
FT
FCsp
spsp
sp
FCsp
CC FT
CTFC
FCsp
TC
sp
spsp
FCsp
DCsp
sp
S1
RC CCsp
DC
LC
sp
1CATA B I
1 2 BPI C I P
2I E API
Bequette, Comp Chem Eng, 20:S1583–S1588, 1996
Togkalidou+, AIChE J, 47:160‐168, 2001
Lakerveld+, Org Process R&D, 19:1088‐1100, 2015
Simon+, Org Process R&D, 19:3‐62, 2015
ISPE Biopharmaceutical Manufacturing Conference
4 – 6 December 2017San Francisco, CA
22
AcknowledgmentsHassan Ait HaddouMonica AminPaul BaroneCatie BartlettLisa BradburyDanielle CampJicong CaoDivya ChandraMichael CostaSteve CramerAleksander CvetkovicAmanda Del RosarioSusan DexterJuyai-Ivy DongChaz GoodwineJongyoon HanPeter HanWilliam HancockWilliam HerringtonShan JiangPankaj KarandeSojin KimSung Hee KoRalf KuriyelTaehong KwonRachel LeesonJames LeungLihong LiuKerry LoveAmos LuTimothy Lu
Huaiya MengSylvia MessierNicholas MozdzierzJustin NelsonRussell NewtonWei OuyangSantosh PandeJoel PaulsonPablo Perez-PineraOliver PurcellGK RajuRajeev RamNigel ReuelDaniel SalemKartik ShahDivya ShastryGajendra SinghAnthony SinskeyStacy SpringsAlan StockdaleMichael StranoKyOnese TaylorSteve TimmickJP TrasattiNicholas VecchiarelloAshley Ford VersyptAnnie WangJames WooDi WuAnna Zhang
This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) and SPAWAR Systems Center Pacific (SSC Pacific) under Contract No. N66001‐13‐C‐4025. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Defense Advanced Research Projects Agency (DARPA) and SPAWAR Systems Center Pacific (SSC Pacific).