Dr. Stéphane Buffat, Director of the LAB (Accidentology, Biomechanics and Driver Behavior), Expert Leader Road Safety, Renault Group

Abstract:

1. Automation and safety improvement

Advanced driver-assistance systems (ADAS) and Automated driving cars (AD) bring many expectations for road safety. However, amongst known pitfalls, the ability to react correctly in a safety-relevant situation will require that the driver maintain some situation awareness, for SAE level 2, and, for the higher levels, to be able to rebuild it, in top of regaining control, during the limited amount of time given for the take-over.

2. Shifts in man-machine interface

There is a place for agents to help bring to the right level the occupants' situation awareness. However, previous attempts to engage in a high-level human-agent system have stressed two areas that needed work:

• Agents shall bring adequate contextual information to anchor an appropriate situation awareness. Such contextual information's exact needs and availability have benefitted from recent work in data mining and sensors.

• Driver's intent detection, which defines the relevant context for the next step, is still mainly at the operational level of simple action for the moment. Most of the current work concentrates on a potential shortage of resources such as distraction, sleep or sudden loss of consciousness.

However, there is more evidence to better understand human driver models, primarily through naturalistic driving studies. These results should help complete the traditional viewpoint named safety I, based on correction of errors, with a more proactive Safety II approach based on safe day to day situation handling.

3. Building trust in the agent:

Finally, a question arises about the necessary conditions for considering valid social agents the automated cooperating systems (usually named cobots in industry 4.0).

The evolution of the perceived distance between robots and humans with internal thought need a conceptualization and question the anthropomorphic view to avoid the uncanny valley.

Experimenting with man-machine Teaming could finally unveil the role of the narrative process while building a "social" trust with a Cooperative system based on progressive and adaptive interaction scenarios.