DCL Seminar: Kartik B. Ariyur - The Problem of Autonomy

Decision and Control Lecture Series

Coordinated Science Laboratory

 

“The Problem of Autonomy”

 

Kartik B. Ariyur, Ph.D.

Purdue University

 

Wednesday, October 12, 2016

4:00 p.m. to 5:00 p.m.

CSL Auditorium (B02)

____________________________________________________________________________________________________________________________________________________________________

 

“The Problem of Autonomy” 

 

Abstract:

From a minor blip on our radars a decade ago, autonomous systems loom large everywhere around us—from our smart phones and computers to drones and self-driving cars, we have systems that monitor their conditions, the environment around them, and attain such simple objectives as minimal power consumption, or complex objectives as safety. They accomplish all of these objectives in unstructured environments like human pockets, roads and urban canyons, as opposed to automated systems that work in completely structured or deterministic and modeled environments like manufacturing lines, where geometry, illumination, and levels of various signals are known precisely.

We will raise several questions here in the context of increasing the autonomy of these systems, and provide some of the answers that have arisen from our group’s research. How can they navigate or locate or orient themselves in their environment? How can they use all of the environmental information that already available to their designers? How do we avoid the propagation of uncertainty of environment into the performance of these systems? Are the methods of automation extensible to autonomous systems? When can large data sets or so called Big Data be exploited to yield actionable information? Is security of these systems possible? Have we developed any artificial intelligence yet? How do we quantify the social objectives and human constraints underlying all of these problems?

Bio:

Kartik B. Ariyur has focused on building autonomy into a variety of systems, i.e., ensuring that systems perform with predictable safety or energy efficiency in uncertain operating conditions. His work on extremum seeking control is used in dozens of industries, his pilot filtering algorithms at Qualcomm running in cell phones worldwide, and health monitoring algorithms for gas turbine engines running in 70% of commercial aircraft APUs (auxiliary power units) worldwide, main engines in regional jets, and the Joint Strike Fighter, all enable autonomous operation of engineering systems. His research straddles most engineering disciplines, and is highly cited (with more than 1200 Google citations for his 60+ papers and 2 books) in addition to patents in industry (17 issued and several more pending). His group’s research at Purdue laid out the framework for safe smart grid renewable integration, maturing solar power to the same level as traditional power sources, modernizing traditional navigation, making it possible for large scale UAV autonomous operations, and increasing the energy efficiency of hydraulic machines. He led the Purdue portion of a project to develop guidelines to integrate system health monitoring with adaptive guidance and control for the USAF. Recently, he led a Mellon Foundation funded project integrating engineering and the humanities and social sciences to develop data driven policy. He leads the signal processing effort for lidar based traffic tracking in the TScan project, helping build the first system tracking traffic autonomously. He is on the Conference Editorial Board of the IEEE Control Systems Society, and on the Editorial Board of the International Journal of Adaptive Control and Signal Processing.