He Sun 孙赫
I am a postdoctoral researcher at Caltech CMS, working with Prof. Katie Bouman. My research focuses on computational imaging and adaptive optics, where multidisciplinary ideas, including optics, control, signal processing and machine learning, are applied. I also have strong interests and industry experiences in robotics.
I received my PhD in Mechanical and Aerospace Engineering from Princeton in 2019, advised by Prof. N. Jeremy Kasdin. My PhD thesis is about space telescope's adaptive optics system for high-contrast exoplanet imaging. I previously received my B.Eng. degree in Engineering Mechanics and B.A. degree in Economics from Peking University in 2014.
Optimal Sensing for Computational Imaging
Optimized sensing is important for computational imaging in low-resource environments, when images must be recovered from severely limited measurements. In this project, we work on deep learning methods that jointly optimize the sensor-sampling strategy and the image reconstruction procedure for computational imaging.
Wavefront Sensing and Control (WFSC)
WFSC, or so-called adaptive optics (AO), is a control system that removes the wavefront aberrations in telescope systems. They are very important in the large ground-based telescopes (e.g. Subaru, Keck) and the next-generation space telescopes (e.g. WFIRST, Habex, LUVOIR) to maintain the high contrast created by coronagraph for faint exoplanets imaging. I contribute to this project in terms of both software and hardware. Recent accomplishments include the applications of adaptive control and optimal experiment design to improve the control system's efficiency, and the development of a broadband imager, an high-contrast integral field spectrograph (HCIFS).
Motion Planning for Autonomous Vehicles
Collaborating with research scientists from Mitsubishi Electric Research Laboratories (MERL), we proposed a novel approach to improve the efficiency of motion planning algorithms by learning heuristics using a deep neural network. Using robotic cars, we experimentally compared this new approach with other search-based planning algorithms (e.g. A*, D*) and sampling-based algorithms (e.g. RRT, RRT* and two-stage RRT). This research mainly considers the motion planning for automated parking, where the car moves slow but the enviroment is very complicated. Click here to see particle filter planning demos from MERL.
Postdoctoral Researcher, California Institute of Technology
, Oct 2019 -
Research Intern, Mitsubishi Electric Research Laboratories (MERL)
, Jun-Sep 2018
- Reviewer for J. Astron. Telesc. Instrum. Syst., Journal of Automatica Sinica, ICASSP
- Member of the Priorities Committee (budgeting committee) of Princeton University, 2017-2018
- MAE433 Automatic Control System, Princeton University, Fall 2018
- MAE206 Introduction to Engineering Dynamics, Princeton University, Spring 2018
- MAE341 Space Flight, Princeton University, Fall 2017
- MAE305/MAT391 Mathematics in Engineering, Princeton University, Fall 2016
- Best paper for observation technologies and systems, IEEE Aerospace Conference, 2019
- Britt and Eli Harari Fellowship, Princeton University, 2018
- SEAS Award for Excellence, Princeton University, 2017
- ExxonMobil Scholarship, Peking University, 2013
- Scholarships for Students in Japan, JASSO, 2013
- Boeing Scholarship, Peking University, 2012
Learning a Probabilistic Sensing Strategy for Computational Imaging Design
Adaptive Optics and Spectroscopy
Robot Motion Planning
He Sun, Adrian V. Dalca, Katherine L. Bouman
Under Review, 2020 [code]
Last updated: by He Sun.