Trust model is a topic that first gained interest in organizational studies and then human factors in automation. Thanks to recent advances in human-robot interaction (HRI) and human-autonomy teaming, human trust in robots has gained growing interest among researchers and practitioners. This article focuses on a survey of computational models of human-robot trust and their applications in robotics and robot controls. The motivation is to provide an overview of the state-of-the-art computational methods to quantify trust so as to provide feedback and situational awareness in HRI. Different from other existing survey papers on human-robot trust models, we seek to provide in-depth coverage of the trust model categorization, formulation, and analysis, with a focus on their utilization in robotics and robot controls. The paper starts with a discussion of the difference between human-robot trust with general agent-agent trust, interpersonal trust, and human trust in automation and machines. A list of impacting factors for human-robot trust and different trust measurement approaches, and their corresponding scales are summarized. We then review existing computational human-robot trust models and discuss the pros and cons of each category of models. These include performance-centric algebraic, time-series, Markov decision process (MDP)/Partially Observable MDP (POMDP)-based, Gaussian-based, and dynamic Bayesian network (DBN)-based trust models. Following the summary of each computational human-robot trust model, we examine its utilization in robot control applications, if any. We also enumerate the main limitations and open questions in this field and discuss potential future research directions.

In this paper, we extend the classic passivity theory for a class of nonlinear impulsive multi-dimensional switched systems with both (exponentially) passive and nonpassive subsystems, where the state changes (i.e. state dimensional variations and/or state jumps) may occur at the switching moments. The passivity conditions of such systems are studied by adopting the transition-dependent average dwell time (TDADT) and multiple Lyapunov functions (MLFs). The state changes of such systems are taken into account by introducing the switching time threshold and the switching rate conditions. The proposed methods prove that, with a relaxed quasi-alternative switching signal, nonlinear impulsive multi-dimensional switched systems with both exponentially passive and nonpassive subsystems are (weakly) exponentially passive, and systems with both passive and nonpassive subsystems are (weakly) passive.  Furthermore, the proved passivity properties of such systems are useful to achieve system stabilization by output feedback. The main results are demonstrated by a numerical example.

Symbolic motion planning for robots is the process of specifying and planning robot tasks in a discrete space, then carrying them out in a continuous space in a manner that preserves the discrete-level task specifications. Despite progress in symbolic motion planning, many challenges remain, including addressing scalability for multi-robot systems and improving solutions by incorporating human intelligence. In this article, distributed symbolic motion planning for multi-robot systems is developed to address scalability. More specifically, compositional reasoning approaches are developed to decompose the global planning problem, and atomic propositions for observation, communication, and control are proposed to address inter-robot collision avoidance. To improve solution quality and adaptability, a hypothetical dynamic, quantitative, and probabilistic human-to-robot trust model is developed to aid this decomposition. Furthermore, a trust-based real-time switching framework is proposed to switch between autonomous and manual motion planning for tradeoffs between task safety and efficiency. Deadlock- and livelock-free algorithms are designed to guarantee reachability of goals with a human-in-the-loop. A set of nontrivial multi-robot simulations with direct human inputs and trust evaluation is provided, demonstrating the successful implementation of the trust-based multi-robot symbolic motion planning methods.

A trust-based switching impedance control strategy for human-robot cooperative manipulation is proposed. The robot behavior switches between the proactive and reactive modes based on the estimated trust of human in robot. The robot estimates the human desired trajectory. A history-based probabilistic trust model for human-robot cooperative manipulation tasks is proposed. Trust is modeled as a dynamic Bayesian network. In the proactive mode, the robot estimates the human desired trajectory and plans accordingly while in the reactive mode, the robot only reacts to the human input. A simulation of the trust-based switching impedance control strategy is presented.

Human-robot collaborations (HRC) can be used for object detection in domain search tasks, which integrate human and computer vision to improve accuracy and efficiency. The Bayesian sequential decision-making (BSD) method has been used for task allocation of a robot in search tasks. In this paper, we first provide an explanation to reveal the nature of the BSD approach: it makes decisions based on the expected value criterion, which is proved to be very different from human decision-making behaviors. On the other hand, it has been shown that joint performance of a team will improve if all members share the same decision-making logic. In HRC, since forcing a human to act like a robot is not desired, we propose to modify the BSD approach such that the robot imitates human logic. In particular, regret theory qualitatively models human's rational decision-making behaviors under uncertainty. We propose a holistic framework to measure regret quantitatively, an individual-based parametric model that fits the measurements, and the integration of regret into the BSD method. Furthermore, we design a human-in-the-loop experiment based on the framework to collect enough data points to further elicit requisite functions of regret theory. Our preliminary results match all the properties in regret theory, while the parametric elicited model shows a good fit to the experimental data.

book chapter | bibtex