Kevin Koch

The globally increasing prevalence and incident rates of mental diseases is one of the most serious public health challenges according to the World Health Organization. Today, treatment is based on professional therapies which require a high amount of financial resources and personnel effort, however IT-supported interventions in ubiquitous devices promise help and a new leverage beyond traditional therapies. We identify the car as a space for new treatments since drivers often have time and the environment in the automobile is highly controlled. In-vehicle information systems can reach people in their daily routine and could introduce innovative prevention measures. In this research in progress paper, we address the open question how the car can improve a driver’s affective state while driving. First, we thoroughly describe the design of a study we conducted to motivate other researchers for this topic. Second, we analyse 631 completed interventions collected in a 2-month field study with 10 drivers. First analyses indicate that we can positively influence the short-term affective state of drivers with at least one of ourintervention types. We provide first practical examples of how to reach the masses of everyday drivers.

Despite efforts to reduce them, traffic accidents continue to increase and bypass reduction targets. The costs of traffic accidents are enormous, killing 1.35 million people every year and costing 3% of most countries' GDP. Recent research aims to target interventions at high-accident-density locations, called accident hotspots. New methods and technologies can systematically identify hotspots, but it remains unclear whether hotspots contribute to accident costs as well as volume. This paper investigates the monetary and human costs of accident hotspots. We analyze a dataset of all accidents from 2011 - 2017 in Switzerland. We identify hotspots, then analyze their contributions to traffic accident costs. We find that hotspot accidents are not different in monetary costliness or injury rates from non-hotspot accidents, so hotspots drive costs along with accident volume. However, hotspot accidents are less fatal, so hotspot targeting might not be best for fatalities. If hotspots are reduced to normal road conditions, total monetary costs can be reduced by up to 5% per year as a theoretical upper bound. Targeting the top 10% most frequent, costly, injurious, or deadly hotspots yeilds different results for different cost types, with accident number and monetary cost targets creating the highest reductions overall.

Driver identification has emerged as a vital research field, where both practitioners and researchers investigate the potential of driver identification to enable a personalized driving experience. Within recent years, a selection of studies have reported that individuals could be perfectly identified based on their driving behavior under controlled conditions. However, research investigating the potential of driver identification under naturalistic conditions claim accuracies only marginally higher than random guess. The paper at hand provides a comprehensive summary of the recent work, highlighting the main discrepancies in the design of the machine learning approaches, primarily the window length parameter that was considered. Key findings further indicate that the longitudinal vehicle control information is particularly useful for driver identification, leaving the research gap on the extent to which the lateral vehicle control can be used for reliable identification. Building upon existing work, we provide a novel approach for the design of the window length parameter that provides evidence that reliable driver identification can be achieved with data limited to the steering wheel only. The results and insights in this paper are based on data collected from the largest naturalistic driving study conducted in this field. Overall, a neural network based on GRUs was found to provide better identification performance than traditional methods, increasing the prediction accuracy from under 15\% to over 65\% for 15 drivers. When leveraging the full field study dataset, comprising 72 drivers, the accuracy of identification prediction of the approach improved a random guess approach by a factor of 25.

In recent years, driver behavior analysis has led to countless driver assistance systems. In these systems, earlier detection of a driver’s maneuver intentions offers opportunities to improve driving experience and safety. Especially brake maneuvers are of fundamental importance because they are directly related to the avoidance of potential hazards.Current state-of-the-art brake assistance systems rely on the release speed of accelerator pedal as an indicator whether a brake event is planned. However, this simple and practical algorithm, fails to capture the overall movement pattern of accelerator pedal behaviors and cannot utilize rich information from different vehicle sensors.To address this issue, we propose a novel recurrent neural network architecture for the purpose of brake maneuver prediction. The proposed method exploits the advantages of multiple sensors. Unlike conventional practices where all signals are aggregated to a single neural network, we leverage the confidence of each sensor. We evaluate our approach based on a dataset of 44 drivers, comprising around 500 hours of naturalistic driving data. The evaluation results show that the proposed algorithm outperforms baseline method by large margin.

Excessive alcohol consumption causes disability and death. Digital interventions are promising means to promote behavioral change and thus prevent alcohol-related harm, especially in critical moments such as driving. This requires real-time information on a person’s blood alcohol concentration (BAC). Here, we develop an in-vehicle machine learning system to predict critical BAC levels. Our system leverages driver monitoring cameras mandated in numerous countries worldwide. We evaluate our system with n = 30 participants in an interventional simulator study. Our system reliably detects driving under any alcohol influence (area under the receiver operating characteristic curve [AUROC] 0.88) and driving above the WHO recommended limit of 0.05 g/dL BAC (AUROC 0.79). Model inspection reveals reliance on pathophysiological effects associated with alcohol consumption. To our knowledge, we are the first to rigorously evaluate the use of driver monitoring cameras for detecting drunk driving. Our results highlight the potential of driver monitoring cameras and enable next-generation drunk driver interaction preventing alcohol-related harm.

Different drivers’ states and emotions can affect negatively the driving performance. Recent advances in affective computing now give the opportunity to measure the users’ state or emotions using various sources of data such as physiological signals or voice samples. Conveying biofeedback in the car could help to make roads safer and improve users’ health and mental state during a ride in an autonomous car. This workshop aims at selecting the drivers’ hazardous states and emotions that are crucial to be assessed, as well as how to convey the appropriate biofeedback to the driver, using multimodal interaction in the car.