CSL Decision and Control Group

Back to Listing

DCL Seminar: Girish Chowdhary - Differentially Constrained Machine Learning for Distributed Autonomy

Event Type
Seminar/Symposium
Sponsor
Decision and Control Laboratory, Coordinated Science Laboratory
Location
CSL Auditorium, Room B02
Date
Oct 26, 2016   3:00 pm  
Speaker
Girish Chowdhary, Ph.D. University of Illinois
Contact
Linda Meccoli
E-Mail
lmeccoli@illinois.edu
Phone
217-333-9449
Views
68

Decision and Control Lecture Series

Coordinated Science Laboratory

 

 “Differentially Constrained Machine Learning for Distributed Autonomy”

 

Girish Chowdhary, Ph.D.

University of Illinois

 

Wednesday, October 26, 2016

3:00 p.m. to 4:00 p.m.

CSL Auditorium (B02)

____________________________________________________________________________________________________________________________________________________________________

 

“Differentially Constrained Machine Learning for Distributed Autonomy” 

 

Abstract:

Recent nonparametric machine learning models such as hierarchic Bayesian Nonparametric Processes (BNPs) and Deep Neural Networks have demonstrated significant success in regression, clustering, and classification on datasets consisting of time-invariant quantities. However, these models remain fundamentally limited in their use as analysis or synthesis tools for feedback control of distributed cyber-physical systems because they are not designed to model systems with smooth spatiotemporal evolution. As a result, fundamental results from systems theory, such as controllability, observability, stability, and invariant subspaces, are not directly leveraged in many existing machine learning techniques.

I will demonstrate how the inclusion of differential constraints in the synthesis of the model can be used to derive fundamental limits of feedback control and estimation with distributed sensors and actuators, and used to develop practical algorithms for CPS predictive inference and autonomy in various challenge domains, including agriculture, environment, and autonomous vehicle systems.  For example, I will show how our Evolving-GP model leads to practical algorithms for predictive inference on spatiotemporal phenomena, and fundamental insights on sensor placement. I will also highlight our recent results in co-robotic training of novice human operators using models learned from expert operators of heavy construction equipment

Bio:

Girish Chowdhary is an assistant professor at the University of Illinois at Urbana-Champaign with Agricultural and Biological engineering (primary) Aerospace Engineering (affiliate), and the Beckman Institute (affiliate). He holds a PhD (2010) from Georgia Institute of Technology. He was a postdoc at the Laboratory for Information and Decision Systems (LIDS) of the Massachusetts Institute of Technology, and an assistant professor at Oklahoma State University for three years prior to moving to UIUC. Girish was a member of the team that launched the first UAV project at the German Aerospace Center's (DLR) in 2003, where he worked for three years. Girish's ongoing research interest is in theoretical insights and practical algorithms for adaptive autonomy, with an application focus on field-robotics and Unmanned Aerial Systems. He is the author of over 90 peer reviewed publications. UAS autopilots based on Girish’s work have been utilized on six UASs, including by independent international institutions. Girish is serving as a PI on NSF, AFOSR, DARPA, and NASA grants and is a co-founder in an agricultural autonomy startup. He is the winner of the Air Force Young Investigator Award, AIAA GNC best paper award, and the Aerospace Guidance and Controls Systems Committee Dave Ward Memorial award for significant contributions to adaptive autopilots.

link for robots only