An Estimation Method of Possible Mode Confusion in Human Work with Automated Control Systems

This paper proposes a new method to estimate possible confusions human users may make on the current mode of an automated control system that they are required to work with. The proposed method takes its stand on a philosophy that the users depend on the system's response behaviors to some extent as valuable cues to sense its state, and thus that different modes in which the system exhibits the same response behaviors to a particular user operation are hard to be distinguished by them. All the modes are encoded into vector forms from the perspective of their "input-output relations," and then the distances among the vectors are utilized as indexes to estimate the likelihoods of mode confusions. The idea is validated through the experiments using a driving simulator with Adaptive Cruise Control (ACC) implemented, in which drivers are queried as to the current mode of the ACC system to collect the probability distributions of mode confusion to be compared with what has been estimated. In addition, it is also examined whether purposefully designed input-output relations can contribute to the improvements of the drivers' awareness of modes.