Date of Award

3-21-2019

Document Type

Thesis

Degree Name

Master of Science in Systems Engineering

Department

Department of Systems Engineering and Management

First Advisor

Michael E. Miller, PhD

Abstract

Autonomous systems have gained an expanded presence within the Department of Defense (DoD). Furthermore, the DoD has clearly stated autonomous systems must extend the capabilities of their human operators. Thus, the exploration of strategies for effective pairing of humans and automation supports this vision. Previous research demonstrated that the time at which an automated agent assumes a task for its human teammate, or agent response time (ART), affects human-agent team performance, human engagement, and human workload. However, in this research environment, the time between subsequent tasks appearing to the human-agent team, or inter-arrival time (IAT), remained constant. Variable IAT environments more accurately reflect real-world operational environments. Previous research also maintained ART at a fixed level. Additionally, the effect of human understanding of automated teammate actions on human-agent team performance remains unknown. This thesis attempts to analyze the effect of an agent with adaptive ART that varies based on current IAT on human-agent team performance, human engagement, and human workload. Additionally, it seeks to determine the implication of agent predictability to the human. This thesis explores these issues in three phases. First, a method and development of a variable ART function for use in future phases is presented. Second, a study of a variable ART teammate against a fixed ART teammate highlights the significance of providing detailed agent instruction to the human. Third, analysis of instruction and type of agent teammate across an entire input IAT function and at different IAT levels is conducted. This work establishes key factors for adaptive ART function implementation. Based on specific IAT changes, the current research demonstrates that adaptive ART can boost human-agent team performance and manipulate human engagement. Furthermore, predictability of agent action in variable IAT environments is a desired system attribute.

AFIT Designator

AFIT-ENV-MS-19-M-166

DTIC Accession Number

AD1076839

Share

COinS