Title: Methods and Models for Robots to Support Variable Human Task Performance

 Date: Thursday, January 18th, 2024

Time: 3:00 - 5:00 PM EST

Location: Klaus 1116

Virtual Link: Zoom

Manisha Natarajan

Robotics PhD Student

School of Interactive Computing

Georgia Institute of Technology

Committee:

Dr. Matthew Gombolay (Advisor) - School of Interactive Computing, Georgia Institute of Technology

Dr. Sonia Chernova - School of Interactive Computing, Georgia Institute of Technology

Dr. Karen Feigh- School of Aerospace Engineering, Georgia Institute of Technology

Dr. Henny Admoni - The Robotics Institute, Carnegie Mellon University

Dr. Laurel Riek - Department of Computer Science and Engineering, University of California San Diego  

Abstract:

The increasing deployment of robots in shared workspaces with humans, spanning applications in defense, manufacturing, and urban search-and-rescue (USAR) missions, highlights the need for devising strategies to enhance human-robot team performance. In my work, I seek to optimize this collaboration by identifying factors influencing human reliance on robots and developing computational methods to enable robots to support, share control, and communicate with users effectively.

In my thesis, I first examine how people trust and depend on suboptimal robot agents offering suggestions in a sequential decision-making task. This work is the first to jointly consider the effect of multiple robot, user, and task attributes, allowing for nuanced multi-way comparisons in identifying critical factors influencing user reliance. Next, I develop data-driven and model-based techniques to facilitate robots to infer human behavior and determine how to assist users. In my proposed work, I seek to formulate a novel computational framework to enable bi-directional communication between humans and robots in a challenging domain, where humans must collaborate with robots to accomplish multiple task objectives. Additionally, I aim to develop novel sampling strategies with data-driven evaluation functions and search heuristics to address the increased computational complexity in multi-objective settings.