Supporting Human-Automation Collaboration through Dynamic Function Allocation: The Case of Space Teleoperation.

Abstract

The implementation of automation technologies in complex domains, such as space operations, poses significant challenges for designers. Proper allocation of tasks and functions to human operators and their automated system is critical and needs to be based on a thorough analysis of automation properties and capabilities. In addition, past experience has shown that fixed designs where tasks and functions are assigned on an a-priori basis often lead to unbalanced workload and poor joint-system performance. To better support human-automation collaboration, this dissertation systematically examines the performance effects of level of automation in support of different stages of information processing during routine operations and in case of automation failures through simulation studies, whose results were also confirmed by a meta-analysis. This research also implements and compares the effectiveness of three dynamic function allocation schemes: adaptive (system-controlled), adaptable (user-controlled), and a hybrid approach where the default mode is adaptive but operators are able to override system selections. The research was conducted in the context of space teleoperation, in particular operation of the robotic arm on the International Space Station. The main findings from this work are that 1) high levels of automation result in improved performance during routine scenarios but incur performance costs in cases of automation failure; 2) operators’ choice of the level of automation depends on their manual skills, task difficulty, and the perceived benefits and problems with (higher level of) automation; 3) adaptable automation seems preferable over adaptive automation because of the increased sense of control and the confusion caused by adaptive automation; and 4) a hybrid system, combining adaptive and adaptable automation, is a promising means of supporting enhanced performance and high operator acceptance but it needs to be refined to eliminate some of the disadvantages of its adaptive component. This research contributes to a better understanding of human-automation collaboration and coordination and provides input to models of joint system performance. Its findings inform the design of effective context-sensitive function allocation schemes and associated interfaces. Finally, this work can be expected to increase the safety and efficiency of operations in a wide range of domains through improved automation design.