Skip to content. Skip to navigation
CIM Menus

NCFRN/REPARTI Research Presentation

OPTIMo: Online Probabilistic Trust Inference Model for Asymmetric Human-Robot Collaborations

Anqi Xu, PhD candidate
Centre for Intelligent Machines McGill University

November 18, 2014 at  2:00 PM
McConnell Engineering Room 437

We present OPTIMo: a dynamic Bayesian model for quantifying the degree of trust that a human supervisor has in an autonomous robot "worker", within an asymmetric collaboration context. OPTIMo maintains beliefs over the user’s moment-to-moment latent trust states based on observable factors, arising both from interaction experiences as well as occasional trust queries. A separate model instance is trained on each individual's experiences, resulting in a personalized characterization of an operator’s behaviors and attitudes. We conducted an observational study on a large group of roboticists, and used the resulting dataset to assess OPTIMo's performance and features, as well as to compare against several existing trust models. OPTIMo's unique abilities, at personalizing and accurately inferring the human's trust states in a near real-time manner, paves the way for trust-aware robots that can learn and adapt to seek out for greater trust and greater efficiency within these human-robot teams.