Human Centered Driving Simulator
As soon as automated driving comes into play in a complex scenario with mixed automated and non-automated vehicles –
trust and, as a consequence, acceptance of systems will be the key to create substantial and sustainable market penetration of autonomous driving.
For creating trust and acceptance a human centric approach will not be enough, since this might lead to correct actions and decisions in a technical sense, but does not consider the human expectation. This might lead to mistrust the technical system. Therefore, at VIRTUAL VEHICLE we focus our research on helping our costumers to create HUMAN-LIKE systems, which are systems whose behaviour is understandable and traceable.
The VIRTUAL VEHICLE simulation environment for developing human-like systems
VIRTUAL VEHICLE is currently expanding its simulation environment towards the possibility to simulate new developments for automated driving in a complex environment taking into account the awareness of the driver and other drivers and pedestrians (Fig. 1).
Figure 1: Simulation environment for ADAS functionalities: Objects are linked and assessed in a cloud and results are passed to the individual players in automated driving.
VIRTUAL VEHICLE’s Drive.LAB as enabler for human-like systems
The VIRTUAL VEHICLE Drive.LAB is a laboratory comprising a driving simulator designed for assessing drivers in a complex mixed-traffic driving scenario with other cars, road users and pedestrians. This dedicated driving simulator can be linked to other driving simulators, to pedestrians that are brought into the loop via augmented or virtual reality and can be connected to the VIRTUAL VEHICLE autonomous driving demonstrator car to close the loop to reality in a seamless
Methodological set-up for developing human-like systems
The human centric needs for safety and comfort in a setting, in which drivers will be confronted with other tasks than simply driving the car (working, relaxing or being entertained) will be addressed in accepted systems only via trust (Fig. 2). Human-like systems will be key to create trust and acceptance. Tutoring the driver towards context- and situation-based interaction with automated systems is another central building block.
Figure 2: Drive.LAB is intended to address human-centered needs like safety and comfort, while creating trust and acceptance even in complex traffic scenarios. The tutoring of drivers to interact with automation is another feature that is offered by Drive.LAB.
Three elements have been established by VIRTUAL VEHICLE to enable human-like system development, which are implemented in Drive.LAB (Fig. 3):
- The Driver Digital Twin: human behavioural models that serve as the basis for all control actions in the loop (these models are continuously updated and enhanced according to measurements in the Drive.LAB)
- Fluid Interaction: considering environmental information (weather, road condition, other vehicles, awareness and state of other drivers and pedestrians), own vehicle state and own state as a driver. Fluid interaction finds situation dependent the best possible way to warn the driver or bring him back into the control loop. The fluid interface, like a fluid, surrounds the driver and continuously adapts to his psychophysical “envelope”. It is a multisensory, omnipresent and omnidirectional system that constantly monitors the driver, his activities and attentional levels to update a driver “digital twin” model. In turn, the updates in the model are used to generate and select the proper sensor modality, timing and location to issue signals and communicate naturally with the driver. The characteristics of a fluid interface enable the implementation of three functions: (i) sustained monitoring of driver, passengers and environment, including V2V and V2X
communications; (ii) control management, including transitions of control (take over and hand back) across different levels of automation; (iii) tutoring of native manual drivers towards increasing automation levels.
- AV Instructor: is used to evaluate the performance of self-driving vehicles with respect to driving style in simulated and real environments.