DATE:

Friday, March 8th, 2019

TIME:

3:30 PM

LOCATION:

GMCS-314

SPEAKER:

Jun Chen, Ph.D. (Purdue University 2018), is an Assistant Professor of
Aerospace Engineering Department at San Diego State University. Dr. Chen’s
research area includes dynamics, control, machine learning and artificial
intelligence, particularly in data-driven modeling, control and optimization
for large-scale networked dynamical systems, with applications in mechanical
and aerospace engineering such as air traffic control, traffic flow management,
and autonomous air/ground vehicle systems. He is presently researching the
implementation of artificial intelligence in aerospace engineering and broader
intelligent transportation system. He received a M.S. degree in Aerospace
Engineering from Purdue University and a B.E degree in Aeronautics Engineering
from Beihang University, China. He is a recipient of the Purdue College of
Engineering Outstanding Research Award in 2018.

ABSTRACT:

With the booming of artificial intelligence in a new era, the field of systems
and control has recently been facing newly emerged research in control
and optimization of large-scale networked autonomous systems, most of
which heavily rely on the fidelity of the models and efficient computational
techniques to execute optimized control actions. Meanwhile, the physical
autonomous systems are inherently subject to uncertainties and disturbances.
This talk will present a suite of modeling, optimization, and computation
algorithms and tools that can efficiently handle uncertainties in large-scale
autonomous systems using the national air transportation system as an example.
In this research, inspired by the state-of-art deep learning and artificial
intelligence research, a data-driven approach is developed for a networked
dynamical autonomous system model, which formulates probabilistic constraints
to consider system uncertainties. In order to efficiently solve such a large-scale
chance-constrained problem, a convex approximation approach is developed, which
takes advantages of and addresses the challenges in modern convex optimization
techniques in solving nonlinear stochastic control and optimization problems.
Most important, the algorithm is a distributed one, and a distributed computing
framework is designed to carry out the computation for the massive independent
approximation processes in the convex approximation, which greatly improves the
computational efficiency. This newly developed chance-constrained optimization
algorithm and platform provide rapid optimal solutions with guaranteed service
level of robustness and will be applicable to many other networked autonomous
systems in mechanical and aerospace engineering.

HOST:

Satchi Venkataraman, Aerospace Engineering

DOWNLOAD:

PDF File