Skip to content. Skip to navigation
CIM Menus

Ph.D. Oral Defense

Efficient collaboration with trust-seeking robots

Anqi Xu
Centre for Intelligent Machines McGill University

September 27, 2016 at  9:00 AM
McConnell Engineering Room 437

We are interested in asymmetric human-robot teams adhering to a supervisor-worker relationship, where the human supervisor occasionally takes over control to aid an autonomous robot agent in its given task. Our research aims to increase and maintain efficient collaboration by improving the robot's task performance, decreasing the human's workload, and sustaining high levels of satisfaction. We address this problem through the lens of trust, which is pervasive among such human-robot teams, and is inherently linked to all of the above efficiency facets.
Our contributions revolve around a novel trust-seeking robot framework that augments an arbitrary robot agent with the abilities to sense and react to the human's changing trust state. This framework includes a fluid interaction paradigm that enables non-expert users to train and help robot agents adapt to changing task conditions. We also elaborate on multiple large-scale user studies that investigated factors from the interaction experience which influence the rapidly-changing dynamics of trust. Building upon these empirical insights, we propose a personalized, probabilistic model for inferring the human's moment-to-moment trust state. This trust inference engine extends the two dominant modeling approaches used in the literature, attains greater prediction accuracy compared to several existing techniques, and features the unique ability to update trust beliefs in real time. We further introduce the first-ever realization of robot agents that react in direct response to the human's trust losses and actively work to restore efficient teamwork. Finally, we demonstrate the diverse efficiency gains of these trust-seeking robots through both extensive controlled experiments as well as challenging real-world field deployments with aerial drones, wheeled robots, and autonomous cars.