Trust-Preserved Human-Robot Shared Autonomy enabled by Bayesian Relational Event Modeling

TL;DR

The paper addresses trust deterioration in long-term human-robot shared autonomy by inferring dynamic trust from time-stamped remote interactions using a Bayesian Relational Event Modeling framework. It combines offline grounding of team properties with online trust updates and couples trust with an adaptive autonomy policy guided by the punctuated equilibrium model to maintain trust and performance. In a collaborative SAR user study, the approach improved success rates, reduced human workload, and enhanced perceived collaboration fluency compared with a baseline SA, demonstrating the practical viability of trust-preserved autonomy. The work offers a principled, longitudinal perspective on trust development and repair in remote HRI, with potential impact on scalable, accepted human-robot teams.

Abstract

Shared autonomy functions as a flexible framework that empowers robots to operate across a spectrum of autonomy levels, allowing for efficient task execution with minimal human oversight. However, humans might be intimidated by the autonomous decision-making capabilities of robots due to perceived risks and a lack of trust. This paper proposed a trust-preserved shared autonomy strategy that allows robots to seamlessly adjust their autonomy level, striving to optimize team performance and enhance their acceptance among human collaborators. By enhancing the relational event modeling framework with Bayesian learning techniques, this paper enables dynamic inference of human trust based solely on time-stamped relational events communicated within human-robot teams. Adopting a longitudinal perspective on trust development and calibration in human-robot teams, the proposed trust-preserved shared autonomy strategy warrants robots to actively establish, maintain, and repair human trust, rather than merely passively adapting to it. We validate the effectiveness of the proposed approach through a user study on a human-robot collaborative search and rescue scenario. The objective and subjective evaluations demonstrate its merits on both task execution and user acceptability over the baseline approach that does not consider the preservation of trust.

Figures (9)

• Figure 1: A search and rescue scenario with human-robot team. Robots may acquire information unavailable to human team members, and make strictly rational decisions.
• Figure 2: A human-robot collaborative SAR scenario where the human and multiple robots jointly perform the search and evacuation of victims from an incident area.
• Figure 3: Examples of sufficient statistics for the collaborative SAR scenario. The solid arrows represent past historical events, and the dashed arrows represent the future events to be predicted. $\mu(i,j,k,t)$ measures the frequency of event $(i,j,k)$ in the past event history before $t$.
• Figure 4: Bayesian REM framework. A moving window technique and sequential Bayesian update rule are utilized to update the estimation of team attributes online based on newly observed relational events.
• Figure 5: The coordination between the robot's autonomy level and the human's trust level.