The Daniel Guggenheim School of Aerospace Engineering

is proud to present

Prof. Nisar Ahmed

Aerospace Engineering Sciences Department
University of Colorado Boulder

"Human-Autonomy Collaboration and the Yin-Yang of Uncertain Decision Making"

Thursday, April 20 @ 12:00pm
Guggenheim Building Room 246

Abstract:
The age of autonomy is here. However, since autonomous systems are products of imperfect human engineering designed to make decisions under uncertainty in a complex world, the promise of ‘set-it-and-forget-it’ autonomy is still quite far off. It is therefore unrealistic to expect autonomous systems to operate out of the box “exactly right”. For sufficiently rich tasks that constantly push the technological cutting edge, they will encounter unexpected situations that require reasoning beyond their designed capabilities. As such, intelligent autonomy must not only be able to gather, process, and act on information independently – they should also be cognizant of what they can and cannot accomplish, and know when and how to seek help.

Human-machine interaction is thus clearly a key component of autonomous system design, alongside perception, planning, learning, etc. An autonomous system should ideally enable stakeholders and users to fluidly (re-)delegate tasks to the system, assess information, and help improve operations -- without requiring users to “babysit” it. Yet, the best balance is difficult to achieve in practice, especially if human-machine interaction is treated as an afterthought or “necessary evil”. Humans can also do more than act as band aids for corner cases where autonomous reasoning algorithms haven’t caught up yet.

This talk will highlight novel pathways that have been developed in recent years to enable collaborative human-machine reasoning from the outset in autonomous system design. The focus will be on probabilistic techniques and user-friendly interfaces (e.g. semantic natural language, sketching) that allow autonomous systems to intelligently augment their reasoning with human input. These allow autonomous systems to gracefully “fill in gaps and cut knots” in hard problems, without undermining the roles and capabilities of either autonomy or humans. These methods also point to the importance of communication and uncertainty in shaping the perception and use of machine autonomy. In this light, it is evident that autonomous systems ought to be embraced as products and extensions of human reasoning, rather than as full replacements. 

About Prof. Ahmed:
Nisar Ahmed received his Ph.D. in Mechanical Engineering at Cornell University, Ithaca, NY in January 2012, focusing on Dynamics, Systems and Controls with minors in ECE and Applied Math. From 2012-2013, he was a postdoctoral research associate in the Cornell Autonomous Systems Lab (ASL) with Professor Mark Campbell. He obtained his B.S. in Engineering (summa cum laude) in 2006 from The Cooper Union in his hometown of New York City. He joined the Aerospace Engineering Sciences Department at CU Boulder as an Assistant Professor in 2014.