Reducing conflicts in haptic shared control during curve negotiation

Abstract

Haptic shared control enables continuous interaction between driver and automatic controller by means of torques on the steering wheel that guide towards a reference trajectory. Current haptic shared control systems induce conflicts when the reference trajectory of the automation differs from the drivers own desired trajectory, which can lead to rejection of the automation. This study investigated whether conflicts can be reduced by individualizing the automation. The approach of the study is twofold. First, a Four Design Choices Architecture (FDCA) is implemented that allows for separate tuning of the reference in a haptic shared controller that determines the total torques by separating feedback and feed-forward torques. Second, a driver-model was used to identify the individual steering parameters and determine the corresponding individualized reference in the curved road-sections. The combination of the FDCA controller and individualized reference was evaluated in an experiment using a fixed-based driving simulator. Sixteen subjects participated in a two-day experiment, where the first day was used to collect manual driving data to identify the individual and average steering parameters to construct the references. On the second day participants drove with four conditions: manual control, a conventional shared controller architecture using feedback torques towards a center-line reference, an individual tuned reference with FDCA controller, and an average tuned reference with FDCA controller. The results showed that the driver-model was able to identify differences in individual curve negotiation styles. Both FDCA controllers resulted in higher objective acceptance and lower workload; conflicts in torques and driver torques decreased significantly compared to manual control and the conventional controller. The usefulness and satisfaction scores were only marginally higher for the individualized FDCA compared to the average FDCA. Interestingly, only the individualized FDCA was perceived more satisfactory and useful than the conventional controller and manual control. Thus, the individualized FDCA can reduce conflicts and increase acceptance compared to the conventional shared controller