As vehicle automation technology continues to mature, there is a necessity for robust remote monitoring and intervention features. These are essential for intervening during vehicle malfunctions, challenging road conditions, or in areas that are difficult to navigate. This evolution in the role of the human operator - from a constant driver to an intermittent teleoperator - necessitates the development of suitable interaction interfaces. While some interfaces were suggested, a comparative study is missing. We designed, implemented, and evaluated three interaction concepts (path planning, trajectory guidance, and waypoint guidance) with up to four concurrent requests of automated vehicles in a within-subjects study with N=23 participants. The results showed a clear preference for the path planning concept. It also led to the highest usability but lower satisfaction. With trajectory guidance, the fewest requests were resolved. The study's findings contribute to the ongoing development of HMIs focused on the remote assistance of automated vehicles.

Research on cycling safety has recently gained the attention of the HCI community. While there have been multiple proposals for automated driving features on bikes, we are unaware of a project that systematically aims to translate and evaluate driver assistance systems from the automotive to the bike domain to promote cycling safety in traffic. Thus, we implemented an adaptive cruise control and a lane-keeping/centering system with hard- and software on a motion-based bicycle simulator and investigated their potential in a virtual reality experiment. Based on performance measurements and subjective ratings, results showed significant improvements in technology acceptance, subjective workload, and driving performance regarding the cruise control. In contrast, the lane-centering and lane-keeping features were rated significantly worse than the baseline without such assistance. The paper concludes with a critical reflection on automated driving features for bicycles.

In conditionally automated driving, assertive takeover requests (TORs) have been found to elicit faster responses in the transition from automated to manual control. However, despite this apparent benefit, concerns have been raised that assertive signals can stress the drivers, resulting in worse takeover quality and subsequent driving performance.

Abstract Driving an automated vehicle requires a clear understanding of its automation capabilities and resulting duties on the driver’s side. This is true across all levels of automation but especially so on SAE levels 3 and below, where the driver has an active driving task performance and/or monitoring role. If the automation capabilities and a driver’s understanding of them do not match, misuse can occur, resulting in decreased safety. In this paper, we present the results from a simulator study that investigated driving mode awareness support via ambient lights across automation levels 0, 2, and 3. We found lights in the steering wheel to be useful for momentary and lights below the windshield for permanent indication of automation-relevant information, whereas lights in the footwell showed to have little to no positive effects on driving mode awareness.

In mixed traffic environments, highly automated vehicles (HAV) can potentially be disruptive and a source of hazards due to their non-human driving behavior and a lack of “traditional” communication means (gestures, eye contact, and similar) to resolve issues or otherwise unclear situations. As a result, additional external human-machine interfaces (eHMI) for automated vehicles that replace the now absent human element in communication have been proposed. In this paper, we present the results from a study, in which two versions of a light band eHMI to communicate driving intend of an automated shuttle were evaluated in a real driving environment. We found that the green-red traffic light metaphor and simple animations could improve interaction success in certain aspects. We also found and discuss that the effect of using vs. not using the visual eHMIs was overall lower than expected and that the shuttle’s position and observable driving behavior seemed to play a larger role than anticipated.

With increasing automation capabilities and a push towards full automation in vehicles, mode awareness, i.e., the driver’s awareness of the vehicle’s current automation mode, becomes an important factor. While issues surrounding mode awareness are known, research concerning and design towards mode awareness appears to not yet be a focal point in the automated driving domain. In this paper, we provide a state-of-the art on mode awareness from the related domains of automated driving, aviation, and Human-Robot Interaction. We present a summary of existing mode awareness interface solutions as well as existing techniques and recognized gaps concerning mode awareness. We found that existing interfaces are often simple, sometimes outdated, yet are difficult to meaningfully expand without overloading the user. We also found predictive approaches as a promising strategy to lessen the need for mode awareness via separate indicators.

Driverless shuttles bear different and novel challenges for passengers. One of these is related to capacity management, as such shuttles are often smaller (usually from 6 to 12 seats) with limited capacities to (re-)assign seating, control reservations, or arrange travels for groups that exceed a shuttle’s capacity. Since a bus driver is missing, passengers need to resolve conflicts or uncertainties on their own, unless additional systems provide such support. In this paper, we present the results from a laboratory study, in which we investigated passenger needs in relation to booking and reserving spots (seats, standing spots, and strollers) in an automated shuttle. We found that such functionalities have a low-to-medium impact on an overall scale but could constitute exclusion criteria for more vulnerable parts of the population, such as older adults, families with small children, or physically impaired individuals.

Abstract The absence of a human driver creates novel challenges for fully automated public transport. Passengers are likely to have different expectations, needs, or even fears when traveling without a driver in potentially dangerous situations. We present the results from two field studies in which we explored incident management in a driverless shuttle bus. We explored participant’s behavior and willingness to assist in solving problems in a variety of scenarios where the bus suddenly stops for technical reasons or a hypothesized situation of harassment. In a follow-up study, we focused on auditory remote assistance and investigated problem solving through the passengers. We found that diffusion of responsibility is an existent barrier, when passengers are involved in the resolving of potentially dangerous situations. It can be overcome, when incident-relevant instructions are designed explicitly, briefly, timely, distinguishable from regular on-trip information, and address auditory and visual sensory channels alike.

Riding a highly automated bus has the potential to bring about a set of novel challenges for the passenger. As there is no human driver present, there is no one to talk to regarding driving direction, stops, or delays. This lack of a human element is likely to cause a stronger reliance on the in-vehicle means of communication, such as displays. In this paper, we present the results from a qualitative study, in which we tested three different on-screen visualizations for passenger information during an automated bus trip. The designs focused primarily on signaling the next stop and proper time to request the bus to stop in absence of a human driver. We found that adding geo-spatial details can easily confuse more than help and that the absence of a human driver makes passengers feel more insecure about being able to exit at the right stop. Thus, passengers are less receptive for visual cues signaling upcoming stops and more likely to input stop requests immediately upon leaving the station.

In light of recent incidents, it has become increasingly relevant to determine who is responsible in case of accidents involving automated vehicles. In this paper, we investigate the question of liability in automated vehicles of SAE levels 3 and above. We claim that there is a mismatch between current liability practices, where a designated driver is usually held responsible, and future perspectives, where the human assumes more and more a passive passenger-like role. Our claims are supported by the results from an interview study with insurance companies from two European countries. We show that insurers lack sufficient data to make informed decisions on how to apportion liability in SAE level 3+ scenarios. We discuss how these considerations have to be reflected in interfaces for the driver in order to make the legal status transparent for the driver.

Automated vehicles have to make decisions, such as driving maneuvers or rerouting, based on environment data and decision algorithms. There is a question whether ethical aspects should be considered in these algorithms. When all available decisions within a situation have fatal consequences, this leads to a dilemma. Contemporary discourse surrounding this issue is dominated by the trolley problem, a specific version of such a dilemma. Based on an outline of its origins, we discuss the trolley problem and its viability to help solve the questions regarding ethical decision making in automated vehicles. We show that the trolley problem serves several important functions but is an ill-suited benchmark for the success or failure of an automated algorithm. We argue that research and design should focus on avoiding trolley-like problems at all rather than trying to solve an unsolvable dilemma and discuss alternative approaches on how to feasibly address ethical issues in automated agents.

Designing safe and effective systems for control transitions between human and vehicle is a difficult task, due to increased reaction times and potentially inattentive drivers. In order to respond to these difficulties, this paper presents an overview of interaction solutions for control transitions between manual and autonomous driving modes. The paper examines technology patents, as well as academic publications. The paper's first contribution is an examination of the current state of the art of control transition interfaces in automated vehicles. The paper's second contribution is the reusable categorization framework developed for this overview. The results are used to identify holes and potentials regarding control transition design, including strong focus on the system over the human, lacking fallback performance, and the potentials of effective driving mode communication. These aspects point the way towards the challenges to be solved -- together with how they might be solved -- for safe and effective control transitions.

In aviation, pilots interact with autopilots almost on a daily basis. With semi-autonomous vehicles, this is not yet the case. In our work, we aimed at finding out what we can learn from pilots' current experiences for the domain of autonomous driving and what implications can be derived. We conducted three in-depth interviews with pilots to investigate how pilots currently handle handover situations to and from the autopilot, which information is relevant for this transition to be successful, how pilots react in critical situations, how handovers are trained, and how flying and handover skills are maintained. We compare the gained insights with the domain of autonomous driving and reflect on implications for handovers and (de)skilling. Our findings suggest that the AUI community can learn from aviation in areas such as situation awareness, transparency of system status, the need for a primary drive display, calibrated (dis)trust, and driver training.

We present active corners, a novel interaction approach for in-car collaboration that maps the spatial arrangement of a car with up to four people (i.e., driver, front seat passenger, and two rear seat passengers) to the form factor of a tablet (i.e., having 4 corners). We use this mental model to allow the driver and passengers to collaborate with each other. Interaction is done by dragging-and-dropping items into the according corners of individual tablets. We present the basic interaction design concept along with three application scenarios. In order to evaluate active corners, we have implemented a multiplayer card game that exploits the proposed interaction concept and fosters collaboration. We report on four field studies in which we evaluated different aspects of the active corners concept. Results show that the interaction approach is intuitive and provides a high usability even in driving conditions.

The distraction potential of communication systems in the automotive context necessitates hands-free and attention undemanding systems. Today's hands-free car kits are of increasingly high quality, since bad audio quality can negatively impact the overall communication quality. Most solutions use built-in speakers for output and a microphone near the driver (e.g. on the ceiling). Thereby, audio quality can suffer e.g. from the long distance between the speaker and the listener. In a recent study, we compared perceived voice quality and social presence of a prototype with speakers installed in the headrest of a vehicle, to a high-end on-board audio system in a communication situation between a person sitting in the driver's seat and a person outside the vehicle. We found that Personal Audio received generally better results while also introducing its own set of issues, e.g., causing spatial disorientation in communication situations, in which other individuals are present in the car.

In the scope of autonomous driving, the question arises if the increased use of automated systems will have an impact on driver's skills in handling the car in the long term. In order to gain more insights on the issue of driver deskilling and how it relates to driving experience and time intervals of non-driving, we conducted an online survey (n=703) considering three driver groups. We found that initial skilling is more of an issue than deskilling after long periods of driving inactivity, i.e., while once learned driving skills seem to remain stable after longer periods of non-driving, they are much more influenced by driving experience in terms of annual mileage and frequency of use. Applied to the autonomous context, this means that drivers must be trained to a high enough skill level or require sufficient manual driving experience, in order to be able to react properly when driving themselves.