Exploring Factors Influencing Human-Autonomy Teams in Automated Driving System Operations

Abstract

Advancing our understanding of human users’ and autonomous systems’ interaction dynamics is critical for the safe operation of complex sociotechnical systems. Defining, modeling, and assessing operational risks arising from human-autonomy interactions are key to evaluating their overall safety. Human-autonomy teams (HATs) emerge as a fundamental concept toward safer designing and evaluating teamwork-related failures and their effect on system-level safety, aiming to evolve from isolated safety analyses toward a holistic perspective of human and machine agents. The adoption of automated driving systems will have a major impact on the transportation environment, where humans will continue to participate as drivers, remote operators, or fellow road users. This article draws from human reliability analysis as well as the cognitive model information, decision, and action in crew context to explore contributing factors shaping tasks and errors across different HATs in automated driving contexts. The impact of organization, agent, scenario, team and other relevant contributing factors on the joint performance of on-board drivers and conditionally automated vehicles are discussed. A candidate selection of factors are proposed to construct a causal HAT model seeking to characterize elements at both design and operation-time performance and for future experimental validation.