News & Event

Seminar

Home > News & Event > Seminar

News & Event

Seminar

Control of Jump Stochastic Systems via Characterization of Their Recurrent Patterns 

Subject

Control of Jump Stochastic Systems via Characterization of Their Recurrent Patterns 

Date

October 5 (Wednesday) 4:30PM

Speaker

SooJean Han, PhD Candidate,  Caltech

Place

E3-2, Room 2203 (제7강의실)

Overview:

O Speaker: SooJean Han, PhD Candidate,  Caltech

O Title: Control of Jump Stochastic Systems via Characterization of Their Recurrent Patterns 

O Date: October 5 (Wednesday) 

O Start Time: 4:30PM

O Venue: E3-2, Room 2203 (제7강의실)

OAbstract:

In this talk, we consider the stability of nonlinear systems with Lévy noise and the control of discrete-time Markovian jump systems. First, we characterize incremental stability for nonlinear systems perturbed by compound Poisson shot noise and finite-measure Lévy noise. Second, for discrete-time Markovian jump systems, we design a controller framework which learns patterns in the jump process to reduce redundant controller synthesis. We demonstrate our approaches with applications such as fault-tolerant control of networks with time-varying topology. 

O Bio:

SooJean Han is a Ph.D. candidate in Control and Dynamical Systems at Caltech under the supervision of Professor Soon-Jo Chung and Professor John C. Doyle. She received her B.S. degree in Electrical Engineering and Computer Science, and Applied Mathematics at UC Berkeley in 2016. In 2017, she was a research assistant in the Hybrid Systems Lab at UC Berkeley, and a research assistant in the Caltech CAST Lab. During an internship at JPL in 2018, she was a part of Team CoSTAR for the DARPA subT Challenge. She received the Caltech Special EAS Fellowship and the NSF GRFP. Her research interests lie in efficient control for stochastic systems that have recurrent jump phenomena, with applications in large-scale networks, fault-tolerance, and aerospace robotics (e.g., swarm intelligence). She is also interested in learning-based control algorithms that take advantage of a system’s patterns over space (e.g., topological symmetries) and over time, such as distributed network control and graph convolutional neural networks. 
 
 

 

Profile: