Automated fault-management in a simulated spaceflight micro-world

• Published in: Aviation, space, and environmental medicine Vol. 73; no. 9; p. 886
• Main Authors: Lorenz, Bernd, Di Nocera, Francesco, Röttger, Stefan, Parasuraman, Raja
• Format: Journal Article
• Language: English
• Published: United States 01.09.2002
• Subjects: Adult; Analysis of Variance; Artificial Intelligence; Automation; Complement Factor H; Female; Humans; Life Support Systems - instrumentation; Male; Space Flight - instrumentation; Spacecraft; NASA Discipline Space Human Factors; Non-NASA Center
• ISSN: 0095-6562

As human spaceflight missions extend in duration and distance from Earth, a self-sufficient crew will bear far greater onboard responsibility and authority for mission success. This will increase the need for automated fault management (FM). Human factors issues in the use of such systems include maintenance of cognitive skill, situational awareness (SA), trust in automation, and workload. This study examine the human performance consequences of operator use of intelligent FM support in interaction with an autonomous, space-related, atmospheric control system. An expert system representing a model-base reasoning agent supported operators at a low level of automation (LOA) by a computerized fault finding guide, at a medium LOA by an automated diagnosis and recovery advisory, and at a high LOA by automate diagnosis and recovery implementation, subject to operator approval or veto. Ten percent of the experimental trials involved complete failure of FM support. Benefits of automation were reflected in more accurate diagnoses, shorter fault identification time, and reduced subjective operator workload. Unexpectedly, fault identification times deteriorated more at the medium than at the high LOA during automation failure. Analyses of information sampling behavior showed that offloading operators from recovery implementation during reliable automation enabled operators at high LOA to engage in fault assessment activities The potential threat to SA imposed by high-level automation, in which decision advisories are automatically generated, need not inevitably be counteracted by choosing a lower LOA. Instead, freeing operator cognitive resources by automatic implementation of recover plans at a higher LOA can promote better fault comprehension, so long as the automation interface is designed to support efficient information sampling.