Quantitative Measures of Regret and Trust in Human-Robot Collaboration Systems

Human-robot collaboration (HRC) systems integrate the strengths of both humans and robots to improve the joint system performance. In this thesis, we focus on social human-robot interaction (sHRI) factors and in particular regret and trust. Humans experience regret during decision-making under uncertainty when they feel that a better result could be obtained if chosen differently. A framework to quantitatively measure regret is proposed in this thesis. We embed quantitative regret analysis into Bayesian sequential decision-making (BSD) algorithms for HRC shared vision tasks in both domain search and assembly tasks. The BSD method has been used for robot decision-making tasks, which however is proved to be very different from human decision-making patterns. Instead, regret theory qualitatively models human’s rational decision-making behaviors under uncertainty. Moreover, it has been shown that joint performance of a team will improve if all members share the same decision-making logic. Trust plays a critical role in determining the level of a human’s acceptance and hence utilization of a robot. A dynamic network based trust model combing the time series trust model is first implemented in a multi-robot motion planning task with a human-in-the-loop. However, in this model, the trust estimates for each robot is independent, which fails to model the correlative trust in multi-robot collaboration. To address this issue, the above model is extended to interdependent multi-robot Dynamic Bayesian Networks.