Conduite complètement automatisée : acceptabilité, confiance et apprentissage de la reprise de contrôle manuel

摘要

Fully automated cars could possibly be on the road in the decades to come. They will allow drivers to be driven by an informatics system in their own vehicle. Such an innovation could lead to a revolution that would change the driver’s status and its activities during the trips, but also the infrastructure, freight, some professions, etc. Nowadays, these vehicles are not available for sale yet, and it is difficult to forecast accurately when they will be, and also what their features will be. Considering this, one of the aims of the present thesis is to examine to what extend fully automated driving will be accepted. Even though the driver is driven by its vehicle, he could have to resume manual control in different circumstances. Indeed, this maneuver could be performed in an emergency or in an anticipated situation while he could be engaged in a non-driving related activity. Performing a manual control recovery could be more or less difficult according to the situation and the experience with the fully automated system. The way this maneuver could be learned by drivers has been examined, testing the impact of different kinds of training on performance and safety (response time and control recovery quality). Acceptability, trust, drivers’ attitudes, intentions to use the fully automated driving system and the impact of these variables on behaviors inside the vehicle have been assessed.Three empirical studies have been realized. The first one, based on two pilot studies, is an online survey (N= 421, 36% of males) assessing attitudes toward fully automated driving, a priori acceptability and intentions of use. This specific driving could require new skills to perform manual control recovery. Hence, learning has become an important point of the second study problematic. This second study (N= 69, 54% of males) has been conducted in a driving simulator on a highway. Attitudes have been measured before and after interacting with this technology. Drivers’ behaviors during the interaction with the fully automated driving system have also been examined. The effect of two kinds of practice (simple vs. elaborated) of fully automated driving on manual control recovery (response times) have been analyzed. Several manual control recovery situations have been tested: some of them were unexpected and urgent, some of them were anticipated. Results showed that the effect of overtrust, increasing response time in the emergency situation, could be moderated for drivers who followed the elaborated practice condition. The aims of the third study are to increase the knowledge one the way to learn how to use fully automated driving, most notably manual control recovery, and to examine the impact of a non-driving related task performing on that maneuver in an emergency situation. Participants (N= 113, 49% of males) have been distributed among two training conditions (simple vs. elaborate), and among two non-driving related task performing conditions (with vs. without). The simple training condition involves a short practice of fully automated driving. The elaborate training condition involves the explanation of the underlying features of the system with a text, a demonstration video, and a practice with more interactions than the simple training condition. After a 20 minutes run, trust, acceptability, attitudes and intentions to use the system have been measured. Elaborated training improved emergency manual control recovery in case of emergency (shorter response times, and greater pedal-use accuracy). Performing a non-driving related task increased response times, but did not significantly impact pedal-use accuracy. Eventually, elaborated training contributed to moderate overtrust in the system.Results of these three studies are discussed. The training methods that have been developed could help training the new users of that technology. Driver status in this kind of vehicle, as well as the nature of the driving task, are debated. Finally, the limits of the studies and perspectives for future research in the field are examined.