TY - GEN
T1 - Identification of the focal plane wavefront control system using E-M algorithm
AU - Sun, He
AU - Kasdin, N. Jeremy
AU - Vanderbei, Robert
N1 - Funding Information:
This work was performed under contract to the Jet Propulsion Laboratory of the California Institute of Technology, award number AWD1004079. We would also like to thank to Samuel Otto from Princeton University for the helpful discussions regarding the E-M algorithm.
Publisher Copyright:
© 2017 COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.
PY - 2017
Y1 - 2017
N2 - In a typical focal plane wavefront control (FPWC) system, such as the adaptive optics system of NASA's WFIRST mission, the efficient controllers and estimators in use are usually model-based. As a result, the modeling accuracy of the system influences the ultimate performance of the control and estimation. Currently, a linear state space model is used and calculated based on lab measurements using Fourier optics. Although the physical model is clearly defined, it is usually biased due to incorrect distance measurements, imperfect diagnoses of the optical aberrations, and our lack of knowledge of the deformable mirrors (actuator gains and influence functions). In this paper, we present a new approach for measuring/estimating the linear state space model of a FPWC system using the expectation-maximization (E-M) algorithm. Simulation and lab results in the Princeton's High Contrast Imaging Lab (HCIL) show that the E-M algorithm can well handle both the amplitude and phase errors and accurately recover the system. Using the recovered state space model, the controller creates dark holes with faster speed. The final accuracy of the model depends on the amount of data used for learning.
AB - In a typical focal plane wavefront control (FPWC) system, such as the adaptive optics system of NASA's WFIRST mission, the efficient controllers and estimators in use are usually model-based. As a result, the modeling accuracy of the system influences the ultimate performance of the control and estimation. Currently, a linear state space model is used and calculated based on lab measurements using Fourier optics. Although the physical model is clearly defined, it is usually biased due to incorrect distance measurements, imperfect diagnoses of the optical aberrations, and our lack of knowledge of the deformable mirrors (actuator gains and influence functions). In this paper, we present a new approach for measuring/estimating the linear state space model of a FPWC system using the expectation-maximization (E-M) algorithm. Simulation and lab results in the Princeton's High Contrast Imaging Lab (HCIL) show that the E-M algorithm can well handle both the amplitude and phase errors and accurately recover the system. Using the recovered state space model, the controller creates dark holes with faster speed. The final accuracy of the model depends on the amount of data used for learning.
KW - E-M algorithm
KW - Exoplanet direct imaging
KW - coronagraph
KW - statistical learning
KW - system identification
KW - wavefront correction
UR - http://www.scopus.com/inward/record.url?scp=85040178876&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85040178876&partnerID=8YFLogxK
U2 - 10.1117/12.2274835
DO - 10.1117/12.2274835
M3 - Conference contribution
AN - SCOPUS:85040178876
T3 - Proceedings of SPIE - The International Society for Optical Engineering
BT - Techniques and Instrumentation for Detection of Exoplanets VIII
A2 - Shaklan, Stuart
PB - SPIE
T2 - Techniques and Instrumentation for Detection of Exoplanets VIII 2017
Y2 - 8 August 2017 through 10 August 2017
ER -