729 research outputs found
Sort by
Computational Modeling of Driver Lateral Control on Curved Roads with Integration of Vehicle Dynamics and Reference Trajectory Tracking
No full textDriver’s lateral control on curved roads plays a significant role in reducing or avoiding the crashes. To understand and predict driver performance on curved roads, a computational model was developed in a cognitive architecture, the Queueing Network-Model Human Processor (QN-MHP), with the integration of vehicle dynamics principles (i.e., how to steer based on near and far angles) and the reference trajectory tracking method (i.e., how to steer on the road varying with radius of road curvature). The model was implemented with four major components: road information, vehicle dynamics, visual perception, and cognition & motor controls. The model outputs were validated with the corresponding human subject performance in the literature. The performance results of the model highly fitted the human subject data such as steering wheel angle
How Do Changes in the External Environment Affect Driving Engagement in Automated Driving? – An Exploratory Study
No full textWe developed a new method for simultaneously assessing theworkload of a driver and a non-driver engaged in natural conversation either inthe vehicle or over a cell phone. For both the driver and non-driver, talking wasfound to be more demanding than listening and the pattern was identical for bothpassenger conversations and cell phone conversations. Operating the vehicleincreased the workload for the driver over and above the conversation task. Theeffects of driving (or not) and talking (or not) were found to be additive. The datareveal a pattern of dynamic fluctuation in workload in driver/non-driverconversational dyads. Driving is performed while processing various internal driver and external cues from the driving environment (e.g., subtle vibrations, lateral and longitudinal acceleration). The present study was conducted for the purpose of identifying how much external cues affect driver’s gaze behavior in an automated driving environment. Fifteen participants drove a commercially available vehicle with longitudinal and lateral automation on an oval test track. Participants were asked to drive the vehicle with and without automation, with or without a side-task, and either with their hands-on or hands-off-wheel. Driver’s gaze behavior, handson-wheel status and driving conditions were annotated from video data. The results showed that during automated driving and side-task performance, eyes-on-road time was significantly greater after entering a curve than before and as a result of changes in speed. These differences were not observed in automated driving mode when no side-task is performed. Also, these were more sensitive than hands-on or hands-off-wheel conditions. The results also suggest that drivers may process nonvisual information (e.g., vestibular information produced by changes in lateral and longitudinal vehicle acceleration) prior to or even during the implementation of a visual resource allocation strategy. The present study suggests driver awareness can be aided without requiring the driver to grab the steering wheel
Effectiveness of Training Interventions on the Hazard Anticipation for Young Drivers Differing in Sensation Seeking Behavior
No full textDrivers younger than 25 years are overrepresented in fatal crashes compared to experienced drivers between 30 and 55 years of age. This age-related difference in crash statistics partly arises from younger drivers’ poor hazard anticipation. Training programs (e.g. SAFE-T; Yamani et al. (2016)) have been shown effective at improving these drivers’ anticipation behavior. However, individual differences such as sensation-seeking behavior, aggression, and driving violations exist in young drivers and may contribute to differences in their hazard anticipation. The current driving simulator study examined whether three individual differences known to characterize driving behavior can predict hazard anticipation performance for young drivers, and training effectiveness. K-mean clustering technique classified participants into two clusters based on their driving aggression, sensation seeking and driving violation scores. The results indicated that the low sensation-seeking drivers benefitted more from the training than their high sensation-seeking peers. These findings have design implications for the development of appropriate countermeasures for high sensation-seeking drivers
Assessing the Distraction Potential of Changeable Highway Message Signs
No full textTwo experiments were conducted to assess how changeable message signs (CMS) within the right-of-way affect driver behavior and attention. Experiment 1 evaluated whether repeated exposure to irrelevant messages would cause drivers to fail to respond to a safety critical message. Experiment 2 evaluated whether the presence of a driving irrelevant message designed to attract attention would cause drivers to fail to respond to a hazard in the roadway. In both experiments, drivers completed a lengthy (about 50 min) driving simulation in a freeway scenario with CMS every 0.8 km (0.5 mi). Dependent measures were gaze location, response to safety critical message (Exp. 1), and response to spilled load in roadway (Exp. 2). It was found that (1) when headways were short, drivers tend to focus on the roadway and not on a CMS; (2) repeated exposure to irrelevant messages did not cause drivers to miss safety critical messages; (3) salient CMS images (changing faces) did not cause failures to detect a roadway hazard, and (4) the frequency and duration of looks to salient images and travel time messages were similar
The Effects of Task Load and Vehicle Heterogeneity on Performance in the Multiple-Vehicle Tracking Task
No full textWhen crossing traffic at busy intersections, drivers must keep track of the changing positions of cyclists, pedestrians and other vehicles to avoid collision. Multiple-object tracking is the ability to monitor the positions of a number of selected moving objects (targets) among others (distractors) in a complex scene. Most young adults can track 3-5 items at once but older adults cannot track as many, a finding that may partially explain older drivers’ increased risk at intersections. Because tracking represents an important component of driving, a variant of the multiple-object tracking task called multiple-vehicle was created to measure tracking performance in a driving simulator. However, it is unclear whether tracking while driving works the same as tracking carried out on its own. Laboratory studies suggest that tracking improves when the moving items are heterogeneous, and on the road, it is far more typical that vehicles differ from one another rather than being all the same. Drivers were given the task of tracking the positions of 4 vehicles in a field of 8 on a highway, and the effects of task load (tracking alone, tracking while driving) on tracking performance were measured as a function of whether the target and distractor vehicles were homogeneous. Steering and headway maintenance variability were also assessed. The results indicated that heterogeneity only enabled better tracking when drivers were tracking in isolation. Heterogeneity had no significant effect on tracking when participants were tracking while driving though it did significantly reduce their steering variability
From Few to Many: Using Copulas and Monte Carlo Simulation to Estimate Safety Consequences
No full textWith the introduction of more advanced vehicle technology, it is paramount to assess its safety benefit. Advanced driver assistance systems (ADAS) can reduce crashes and mitigate crash severity, if designed appropriately. Driver behavior models are integral to the ADAS design process, complementing time and resource intensive human participant experiments. We introduce a method to model driver responses to forward collision events by quantifying multivariate behavior with copulas and Monte Carlo simulation. This approach capitalizes on the data from small samples of crash events observed in naturalistic or simulator studies. Copulas summarize data by capturing the underlying joint distribution of variables, and Monte Carlo methods can be used to repeatedly sample from these distributions. A driver model can be parameterized with these samples, and run on a desktop driving simulation environment
Pilot Results on Forward Collision Warning System Effectiveness in Older Drivers
No full textAdvanced Driver Assistance Systems (ADAS) have largely been developed with a “one-size-fits-all” approach. This approach neglects the large inter-individual variability in perceptual and cognitive abilities that affect aging ADAS users. We investigated the effectiveness of a forward collision warning (FCW) with fixed response parameters in young and older drivers with differing levels of cognitive functioning. Drivers responded to a pedestrian stepping into the driver’s path on a simulated urban road. Behavioral metrics included response times (RT) for pedal controls and two indices of risk penetration (e.g., maximum deceleration and minimum time-to-collision (TTC)). Older drivers showed significantly slower responses at several time points compared to younger drivers. The FCW facilitated response times (RTs) for older and younger drivers. However, older drivers still showed smaller safety gains compared to younger drivers at accelerator pedal release and initial brake application when the FCW was active. No significant differences in risk metrics were observed within the condition studied. The results demonstrate older drivers likely differ from younger drivers using a FCW with a fixed parameter set. Finally, we briefly discuss how future research should examine predictive relationships between domains of cognitive functioning and ADAS responses to develop parameter sets to fit the individual
The Driver Has Control: Exploring Driving Performance with Varying Automation Capabilities
No full textAs vehicle automation becomes more capable and prevalent, an understanding of how drivers will interact with automation systems of varying capabilities will be of critical importance. In this study, we compare the performance of drivers on takeover of control from varying types of automation systems (single-function and combined function). Participants drove a 20-minute course with sections of automated driving, and with several traffic events designed to elicit a driver response. Structured transfers of control between automated and manual driving modes occurred following a 7-second countdown at fixed locations on the course. Significant differences were found between groups in terms of lanekeeping ability immediately after taking control following a period of automated vehicle control or partial driver/automation control, but significant differences were not found in accident evasion ability, even five seconds after resuming full control
Cognitive Distraction Impairs Drivers' Anticipatory Glances: An On-Road Study
No full textThis study assessed the impact of cognitive distraction on drivers’ anticipatory glances. Participants drove an instrumented vehicle and executed a number of secondary tasks associated with increasing levels of mental workload including: listening to the radio or audiobook, talking on a handheld or hands-free cellphone, interacting with a voice-based e-mail/text system, and executing a highly demanding task (Operational Span task; OSPAN). Drivers’ visual scanning behavior was recorded by four different high definition cameras and coded offline frame-by-frame. Visual scanning behavior at road intersections with crosswalks was targeted because distraction is one of the major causes of accidents at these locations (NHTSA, 2010a). Despite the familiarity of the locations, results showed that as the secondary-task became more cognitively demanding drivers reduced the amount of anticipatory glances to potential hazards locations. For example, while interacting with a high fidelity voice-based email/text system, the probability of executing a complete scan of the intersection was reduced by 11% compared to the no-distraction control condition. These results document the effects of cognitive distraction on drivers’ visual scanning for potential hazards and highlight the detrimental role of voice based systems on driving behavior
Developing and Testing Operational Definitions for Functional and Higher Order Driving Instruction
No full textThe amount and type of driving instruction provided to novice teen drivers during the learner period may be associated with future crash risk. The purpose of this study was to (1) operationally define two types of driving instruction: functional and higher order instruction, and to (2) test these definitions in a sample of newly licensed novice teenage drivers during the first ten hours of supervised driving. Functional driving instruction was defined as instruction that relates to the present time or immediate future; and related to specific events that are occurring during the drive itself. Higher order driving instruction was instruction that could be extrapolated to a future driving situation; that conveys general principles of driving related to potential events that occur. These operational definitions were tested in conversation occurring during driving instruction in a sample of 90 teen drivers, recruited within three weeks of receiving their learner permit. Teen drivers’ vehicles were equipped with microphones; conversations were recorded and coded for each type of instruction that was observed. As expected, parents provided substantial driving-related instruction on a variety of topics. During the first ten hours of supervised driving only 17.5% of observed driving-related instructions was higher order. This test provides face validity of the operational definitions of driving instruction. These definitions may assist in quantifying the type and amount of driving instruction occurring during the supervised practice stage of licensure, and provide an empirical basis for evaluating the association between driving instruction and independent driving performance