Context Presentation
This internship concerns autonomous vehicles of level 3. A level 3 vehicle is only capable of taking full control and operating given specific conditions. However the vehicle is not fully autonomous and the human driver must remain vigilant even in the autonomous driving phase as he can be asked at any moment to retake control over the vehicle in a short amount of time in the event of a failure (situation too complex for the vehicle to handle for example). It is therefore important to make sure that the human driver is always in a state fit to take back control.
In this context, our goal is to design a system capable of observing the human driver while using the vehicle in various situations and to extract a personal profile from these observations that will then be used to measure the attention of the human driver. In the case of low attention, the system will also need to act by either trying to increase the attention of the human driver, or telling the autonomous vehicle that the human driver is not fit to take back control at a moment's notice.
Figure 1 : « A distracted driver in an autonomous vehicle »
Scientific Goals
- to understand and define the intrinsic and contextual factors defining the driver's attention level
- to design a multi-agent system to estimate at runtime the human driver's attention level and to maintain it based on contextual information
- to define and implement an experimental protocol to evaluate the proposed multi-agent system.
Keywords
Autonomous vehicle, cooperative multi-agent system, dynamic learning, synthetic environment.
Contacts
michael.geraedts-muse@irit.fr,
