Analyzing the take-over performance in an automated vehicle in terms of cognitive control modes - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Le travail humain Année : 2020

Analyzing the take-over performance in an automated vehicle in terms of cognitive control modes

Résumé

Several studies have pointed out that inter-individual differences could exist in the way drivers interact with an automated vehicle. Some of them revealed that drivers can be classified into different groups according to their behaviour. In line with these studies, the present paper aims at identifying different classes of drivers using clustering methods, according to their behaviour after a Take-Over Request (TOR). It uses the concept of “cognitive control modes” to interpret them.This concept was defined by Hollnagel (1993, 2002) who distinguished different modes of cognitive control, from the most reactive to the most proactive. These modes are associated with different kinds of performance (pattern of actions), which are more or less efficient.This study relies on data collected in a driving simulator experiment, during which 36 participants experienced automated driving at SAE level 3 (SAE J3016, 2018). They were invited to play a game on a tablet placed under the windscreen. The TOR took place in a lane-changing situation. Participants had 10 s to resume control and afterwards had to perform a lane change. The study focused on a condition in which the drivers used an Augmented Reality Head Up Display which aimed to help them build a satisfying awareness of the situation quickly by drawing their attention to the most important features of the driving scene, and to facilitate their understanding. Several kinds of data were considered in order to classify the participants and to explain the resulting classes, these were: drivers’ reactions using in-vehicle data (related to driver control and vehicle movement), eye-tracking data, and verbal data from post-activity interviews.Clustering methods were used to process in-vehicle data. They helped to identify three patterns of data, or “behavioural classes”. Class 1 is related to smooth actions, a positive user experience; participants in this class spent more time looking at the driving scene compared with other classes. Also characterised by smooth actions, Class 2 is related to a more mitigated experience. On the other hand, Class 3 is associated with: abrupt braking actions, a faster lane change, negative user experience; with eye fixations on the game tablet persisting after the TOR, as well as with difficulties in understanding the information displayed.These classes of behaviours have been interpreted in terms of different cognitive control modes: a “tactical” control mode implemented when drivers give themselves enough time to analyse the situation, a “scrambled” or “opportunistic” control mode when they do not.
Fichier non déposé

Dates et versions

hal-03157240 , version 1 (03-03-2021)

Identifiants

Citer

Christine Chauvin, Farida Saïd, Philippe Rauffet, Sabine Langlois. Analyzing the take-over performance in an automated vehicle in terms of cognitive control modes. Le travail humain, 2020, 83 (4), pp.379-405. ⟨10.3917/th.834.0379⟩. ⟨hal-03157240⟩
81 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More