Search

Non-driving-related tasks (NDRTs) have the potential to affect safety in a number of ways, but the conditions under which drivers choose to engage in NDRTs has not been extensively studied. This analysis considers naturalistic driving data in which drivers were recorded driving and engaging in NDRTs at will for several weeks. Using human-annotated video captured from vehicle cabins, we examined the probabilities with which drivers engaged in NDRTs, and we examined the relationship between vehicle speed and NDRT probability, with the goal of modeling NDRT probability as a function of speed and type of NDRT observed. We found that tasks that contain significant visual and manual components, such as phone manipulation, show strong sensitivity to vehicle speed, while other tasks, such as phone conversation, show no effects of vehicle speed. These results suggest that there are systematic relationships between NDRT patterns and vehicle speed, and that the nature of these relationships is sensitive to the demands of the NDRT. The relationship between speed and NDRT probability has implications for understanding the effects of NDRTs on safety, but also for understanding how drivers may differ in terms of the strategies they employ to modulate their NDRT behaviors based upon driving demands.

Today, and possibly for a long time to come, the full driving task is too complex an activity to be fully formalized as a sensing-acting robotics system that can be explicitly solved through model-based and learning-based approaches in order to achieve full unconstrained vehicle autonomy. Localization, mapping, scene perception, vehicle control, trajectory optimization, and higher-level planning decisions associated with autonomous vehicle development remain full of open challenges. This is especially true for unconstrained, real-world operation where the margin of allowable error is extremely small and the number of edge-cases is extremely large. Until these problems are solved, human beings will remain an integral part of the driving task, monitoring the AI system as it performs anywhere from just over 0% to just under 100% of the driving. The governing objectives of the MIT Advanced Vehicle Technology (MIT-AVT) study are to 1) undertake large-scale real-world driving data collection that includes high-definition video to fuel the development of deep learning-based internal and external perception systems; 2) gain a holistic understanding of how human beings interact with vehicle automation technology by integrating video data with vehicle state data, driver characteristics, mental models, and self-reported experiences with technology; and 3) identify how technology and other factors related to automation adoption and use can be improved in ways that save lives. In pursuing these objectives, we have instrumented 23 Tesla Model S and Model X vehicles, 2 Volvo S90 vehicles, 2 Range Rover Evoque, and 2 Cadillac CT6 vehicles for both long-term (over a year per driver) and medium-term (one month per driver) naturalistic driving data collection. Furthermore, we are continually developing new methods for the analysis of the massive-scale dataset collected from the instrumented vehicle fleet. The recorded data streams include IMU, GPS, and CAN messages, and high-definition video streams of the driver's face, the driver cabin, the forward roadway, and the instrument cluster (on select vehicles). The study is on-going and growing. To date, we have 122 participants, 15610 days of participation, 511638 mi, and 7.1 billion video frames. This paper presents the design of the study, the data collection hardware, the processing of the data, and the computer vision algorithms currently being used to extract actionable knowledge from the data.

Forward collision warning (FCW) systems typically employ forward sensing technologies to identify possible forward collisions and provide an alert to the driver in the event they have not recognized a threat. These systems have demonstrated safety benefits. However, because the base rate of collisions is low, sensitive FCW systems can provide a high rate of alarms in situations with no or low probability of collision, which may negatively impact driver responsiveness and satisfaction. We examined over 2000 naturally occurring FCWs in two modern vehicles as a part of a naturalistic driving study investigation into advanced vehicle technologies. Analysts used cabin and forward camera footage, as well as environmental characteristics, to judge the likelihood of a crash during each alert, which were used to model the likelihood of an alert representing a possible collision. Only nine FCWs were considered “crash possible and imminent”. Road-type, speed, traffic density, and deceleration profiles distinguished between alert severity. Modeling outcomes provide clues for reducing nuisance and false alerts, and the method of using subjective video annotation combined with vehicle kinematics shows promise for investigating FCW alerts in the real world.

Accurate, robust, inexpensive gaze tracking in the car can help keep a driver safe by facilitating the more effective study of how to improve (i) vehicle interfaces and (ii) the design of future advanced driver assistance systems. In this study, the authors estimate head pose and eye pose from monocular video using methods developed extensively in prior work and ask two new interesting questions. First, how much better can they classify driver gaze using head and eye pose versus just using head pose? Second, are there individual‐specific gaze strategies that strongly correlate with how much gaze classification improves with the addition of eye pose information? The authors answer these questions by evaluating data drawn from an on‐road study of 40 drivers. The main insight of the study is conveyed through the analogy of an ‘owl’ and ‘lizard’ which describes the degree to which the eyes and the head move when shifting gaze. When the head moves a lot (‘owl’), not much classification improvement is attained by estimating eye pose on top of head pose. On the other hand, when the head stays still and only the eyes move (‘lizard’), classification accuracy increases significantly from adding in eye pose. The authors characterise how that accuracy varies between people, gaze strategies, and gaze regions.

The relationship between a driver’s glance orientation and corresponding head rotation is highly complex due to its nonlinear dependence on the individual, task, and driving context. This paper presents expanded analytic detail and findings from an effort that explored the ability of head pose to serve as an estimator for driver gaze by connecting head rotation data with manually coded gaze region data using both a statistical analysis approach and a predictive (i.e., machine learning) approach. For the latter, classification accuracy increased as visual angles between two glance locations increased. In other words, the greater the shift in gaze, the higher the accuracy of classification. This is an intuitive but important concept that we make explicit through our analysis. The highest accuracy achieved was 83% using the method of Hidden Markov Models (HMM) for the binary gaze classification problem of (a) glances to the forward roadway versus (b) glances to the center stack. Results suggest that although there are individual differences in head-glance correspondence while driving, classifier models based on head-rotation data may be robust to these differences and therefore can serve as reasonable estimators for glance location. The results suggest that driver head pose can be used as a surrogate for eye gaze in several key conditions including the identification of high-eccentricity glances. Inexpensive driver head pose tracking may be a key element in detection systems developed to mitigate driver distraction and inattention.

Previous literature has shown that vehicle crash risks increases as drivers' off-road glance duration increases. Many factors influence drivers' glance duration such as individual differences, driving environment, or task characteristics. Theories and past studies suggest that glance duration increases as the task progresses, but the exact relationship between glance sequence and glance durations is not fully understood. The purpose of this study was to examine the effect of glance sequence on glance duration among drivers completing a visual-manual radio tuning task and an auditory-vocal based multi-modal navigation entry task. Eighty participants drove a vehicle on urban highways while completing radio tuning and navigation entry tasks. Forty participants drove under an experimental protocol that required three button presses followed by rotation of a tuning knob to complete the radio tuning task while the other forty participants completed the task with one less button press. Multiple statistical analyses were conducted to measure the effect of glance sequence on glance duration. Results showed that across both tasks and a variety of statistical tests, glance sequence had inconsistent effects on glance duration-the effects varied according to the number of glances, task type, and data set that was being evaluated. Results suggest that other aspects of the task as well as interface design effect glance duration and should be considered in the context of examining driver attention or lack thereof. All in all, interface design and task characteristics have a more influential impact on glance duration than glance sequence, suggesting that classical design considerations impacting driver attention, such as the size and location of buttons, remain fundamental in designing in-vehicle interfaces.