Authority transitions in highly automated vehicles

Abstract

Relevance -
Autonomous cars can offer many benefits and potentially nullify the road accidents caused by human error (94%). One of the threats, however, is the mode transitions between the levels of automation. In a highly automated vehicle (level 4), the vehicle can drive completely autonomous, but only during a defined use case after which the user has to take over. This leads to sudden changes in workload which can be detrimental to driving safety. The current designs do not acknowledge these safety issues yet. Therefore the goal of this project is to design a system where the transitions to and from autonomous driving in a highly automated car are comfortable and safe. The project is executed in collaboration with the design agency VanBerlo.

Literature study -
A literature study led to the identification of seven human-factors issues that affect the transitions between levels of vehicle automation, expected to emerge between 2020 and 2032: Vigilance decrements, complacency, manual skill decay, motion sickness, loss of situational awareness, predictability of the car and mode transitions
For automated driving, being a new development, it seems recommendable that the design should focus on building trust with the system and creating a built-in co-pilot that supports the driver in the new, more complex cockpit.
These different automation modes have been divided in assisted driving and autonomous driving with the difference being that in the latter one the system is responsible. In assisted driving, automated features like adaptive cruise control (ACC) and lane keeping can still be used making the system more complex. During autonomous driving, three measurable user states have been defined. Different studies show it takes 30-5 minutes from a sleeping state, 40 seconds from an inattentive state and 8 seconds from an attentive state to get back to the right driving performance level.

Field research -
A test ride with a partial-automated vehicle (Tesla model S) shows minimal feedback from the automated features which leads to confusion. Also, trust was a reoccurring issue emphasised during this ride. Observing six students of a driving instructor during the lessons show unsolved issues in planning, traffic rules and communication. Calm, clear and structured feedback from the instructor was favoured. The driving instructor emphasises the importance of communication to other road users.

Design goal -
The research concludes in the design goal: ‘Make the transitions between fully automated and assisted driving safer by increasing the situational awareness when needed, re-engaging driver vigilance, avoiding mode confusion and establish trust with the system.’

Design -
During assisted driving the HUD shows in combination with augmented reality what the status is of ACC and lane keeping and visualises their functionality to prevent confusion. Levers behind the steering wheel are used to control these features. Holding both activates autonomous driving. The steering wheel provides haptic feedback to indicate the take over after which it retracts as a strong symbolic message of giving the steering wheel to the system.
A framework is designed based on the literature research to facilitate the transition back to assisted driving where situational awareness and driver vigilance increase stepwise before a take over. Haptic feedback in the seat emphasises and pushes the user in the desired user state. If the desired user state is not met afterwards, auditive feedback is used to warn the user until the user state is met. These modalities are always combined with visual feedback to make the message clear. Holding both levers again initiated assisted driving. The steering wheel comes back as a symbolic message of taking the wheel and provides haptic feedback at the point of take over.

Evaluation -
A qualitative study with an interactive prototype, potential users and experts on automated driving is done to evaluate the design. It shows that the feedback modalities of the design work to facilitate the transitions. It is expected, however, that the steering wheel is presented back to the user in an earlier stage to initiate the take over. It remains unclear how the augmented elements work to facilitate ACC and lane keeping.
Next steps should be to evaluate the design further and design a system that communicates with other road users and add more functionality to the retracting steering wheel.