Title:Assessing Drivers' Situation Awareness in Semi-Autonomous Vehicles: ASP based Characterisations of Driving Dynamics for Modelling Scene Interpretation and Projection

Abstract:Semi-autonomous driving, as it is already available today and will eventually become even more accessible, implies the need for driver and automation system to reliably work together in order to ensure safe driving. A particular challenge in this endeavour are situations in which the vehicle's automation is no longer able to drive and is thus requesting the human to take over. In these situations the driver has to quickly build awareness for the traffic situation to be able to take over control and safely drive the car. Within this context we present a software and hardware framework to asses how aware the driver is about the situation and to provide human-centred assistance to help in building situation awareness. The framework is developed as a modular system within the Robot Operating System (ROS) with modules for sensing the environment and the driver state, modelling the driver's situation awareness, and for guiding the driver's attention using specialized Human Machine Interfaces (HMIs).
A particular focus of this paper is on an Answer Set Programming (ASP) based approach for modelling and reasoning about the driver's interpretation and projection of the scene. This is based on scene data, as well as eye-tracking data reflecting the scene elements observed by the driver. We present the overall application and discuss the role of semantic reasoning and modelling cognitive functions based on logic programming in such applications. Furthermore we present the ASP approach for interpretation and projection of the driver's situation awareness and its integration within the overall system in the context of a real-world use-case in simulated as well as in real driving.