Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Event-Triggered Framework for Trust-Mediated Human-Autonomy Interaction

Daniel A. Williams, Airlie Chapman, Chris Manzie

TL;DR

The paper addresses the challenge of modeling and ensuring stable operation in human-on-the-loop human–autonomy interaction by proposing a general hybrid framework that integrates trust dynamics, commander interventions, and autonomous system behavior. It introduces five interconnected subsystems (Commander Intervention Interface, System Controller and Dynamics, System Status Interface, Performance Estimation, and Commander Trust and Intervention Dynamics) and analyzes their interactions under asynchronous event-triggered sampling, proving uniform global asymptotic stability of the closed-loop set ${oldsymbol{ ext{A}}_p imes oldsymbol{ ext{A}}_c}$ via a Lyapunov-based small-gain approach. The framework is demonstrated through a swarm search-and-rescue numerical example, showing how tuning sampling intervals and controller gains influences convergence, sampling frequency, and cognitive load. The results highlight the practical potential of event-triggered, trust-aware interfaces to improve efficiency and safety in human-autonomy teams while maintaining rigorous stability guarantees.

Abstract

Inspired by the increased cooperation between humans and autonomous systems, we present a new hybrid systems framework capturing the interconnected dynamics underlying these interactions. The framework accommodates models arising from both the autonomous systems and cognitive psychology literature in order to represent key elements such as human trust in the autonomous system. The intermittent nature of human interactions are incorporated by asynchronous event-triggered sampling at the framework's human-autonomous system interfaces. We illustrate important considerations for tuning framework parameters by investigating a practical application to an autonomous robotic swarm search and rescue scenario. In this way, we demonstrate how the proposed framework may assist in designing more efficient and effective interactions between humans and autonomous systems.

An Event-Triggered Framework for Trust-Mediated Human-Autonomy Interaction

TL;DR

The paper addresses the challenge of modeling and ensuring stable operation in human-on-the-loop human–autonomy interaction by proposing a general hybrid framework that integrates trust dynamics, commander interventions, and autonomous system behavior. It introduces five interconnected subsystems (Commander Intervention Interface, System Controller and Dynamics, System Status Interface, Performance Estimation, and Commander Trust and Intervention Dynamics) and analyzes their interactions under asynchronous event-triggered sampling, proving uniform global asymptotic stability of the closed-loop set via a Lyapunov-based small-gain approach. The framework is demonstrated through a swarm search-and-rescue numerical example, showing how tuning sampling intervals and controller gains influences convergence, sampling frequency, and cognitive load. The results highlight the practical potential of event-triggered, trust-aware interfaces to improve efficiency and safety in human-autonomy teams while maintaining rigorous stability guarantees.

Abstract

Inspired by the increased cooperation between humans and autonomous systems, we present a new hybrid systems framework capturing the interconnected dynamics underlying these interactions. The framework accommodates models arising from both the autonomous systems and cognitive psychology literature in order to represent key elements such as human trust in the autonomous system. The intermittent nature of human interactions are incorporated by asynchronous event-triggered sampling at the framework's human-autonomous system interfaces. We illustrate important considerations for tuning framework parameters by investigating a practical application to an autonomous robotic swarm search and rescue scenario. In this way, we demonstrate how the proposed framework may assist in designing more efficient and effective interactions between humans and autonomous systems.

Paper Structure

This paper contains 20 sections, 1 theorem, 26 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. HSI Trust Framework
  3. Commander Intervention Interface
  4. System Controller and Dynamics
  5. System Status Interface
  6. Performance Estimation
  7. Commander Trust and Intervention Dynamics
  8. State Convergence under Event-Triggered Sampling
  9. Plant Dynamics under ETS
  10. Controller Dynamics under ETS
  11. Closed-Loop Dynamics under ETS
  12. Numerical Example
  13. Implementation
  14. Commander Intervention Interface
  15. System Controller and Dynamics
  16. ...and 5 more sections

Key Result

Theorem 1

Given the system eq:sysflows--eq:sysjumps under Assumptions asm:cii_osclb--asm:smallgain, if there exists a tuple of positive constants $(\tau^*_p,\tau^*_c)$ such that for all $\tau_p\in(0,\tau^*_p)$ and $\tau_c\in(0,\tau^*_c),$$G(D)\subset C\cup D$ and $F(q)\in T_C(q)$ for any $q\in C\backslash D$,

Figures (7)

  • Figure 1: Overview of the proposed HAI framework.
  • Figure 2: Swarm centroid trajectory from $t = 0.0$ s to $t=40.0$ s in Mission $M^C_1$. Legend: ★ survivor detected from $t=32.5$ s to $t=40.0$ s, ★ survivor detected before $32.5$ s, ☆ undetected survivor, swarm centroid position when the adjoined survivor was detected, $\circ$ swarm centroid initial position, $-\,-$ trajectory before $t=32.5$ s, --- trajectory from $t=32.5$ s to $t=40.0$ s, ✳ agent positions at $t=40.0$ s.
  • Figure 3: State trajectories in Missions $M^A_{1}$ (blue dashed), $M^B_{1}$ (yellow dash-dotted), $M^C_{1}$ (red dotted) and $P^*$ or $T^*$ (black solid).
  • Figure 4: State trajectories in Missions $M^A_{0.1}$ (blue dashed), $M^B_{0.1}$ (yellow dash-dotted), $M^C_{0.1}$ (red dotted) and $P^*$ or $T^*$ (black solid).
  • Figure 5: Plant sampler state residual during first 10 seconds of $M_1^A$ for $k_p =4$ (black) and $k_p=40$ (blue).
  • ...and 2 more figures

Theorems & Definitions (11)

  • Remark 1
  • Remark 2
  • Remark 3
  • Remark 4
  • Remark 5
  • Remark 6
  • Definition 1
  • Definition 2
  • Remark 7
  • Theorem 1
  • ...and 1 more