Dev S. Kochhar

Driver Distraction: Evaluation with Event Detection Paradigm

The effects of eight in-vehicle tasks on driver distraction were measured in a large, moving-base driving simulator. Forty-eight adults, ranging in age from 35 to 66, and 15 teenagers participated in the simulated drive. Hand-held and hands-free versions of phone dialing, voicemail retrieval, and incoming calls represented six of the eight tasks. Manual radio tuning and climate control adjustment were also included to allow comparison with tasks that have traditionally been present in vehicles. During the drive the participants were asked to respond to sudden movements in surrounding traffic. The driver’s ability to detect these sudden movements or events changed with the nature of the in-vehicle tasks that were being performed. Driving performance measures such as lane violations and heading error were also computed. The performance of the adult group was compared with the performance of the teenage drivers. Compared with the adults, the teens were found to choose unsafe following distances, have poor vehicle control skills, and be more prone to distraction from hand-held phone tasks.

Effects of Adaptive Lane Departure Warning System on Driver Response to a Surprise Event

A lane departure warning (LDW) system monitors the current lane position of a vehicle and presents a driver alert when one of the vehicles front tires crosses a threshold, for example, the nearest lane line. The primary intent of such warning systems is to prevent or mitigate road departures and related crashes caused by driver distraction or drowsiness. The present evaluation compared adaptive and nonadaptive versions of an LDW system. The adaptive version adapted to the drivers state, whereas the nonadaptive version did not. The adaptive LDW system alerted the driver only if a driver state monitor (DSM) indicated that the driver was looking away from the road ahead for 2 s or longer at about the time when a lane line was crossed. Forty volunteers drove a high-fidelity, moving-base driving simulator in a study to compare driver responses to a surprise lane departure when they used a nonadaptive LDW system and then an adaptive LDW system or vice versa. The results indicated that in the adaptive LDW mode, 13 subjects (34%) either experienced delayed activation of the LDW alert or received no LDW alert at all when they should have, primarily because of both the 2-s rule in the adaptive LDW algorithm and DSM registration issues. The adaptive LDW resulted in significantly larger lane excursions at the onset of the LDW alert compared with those that occurred in the non-adaptive LDW mode. These results highlight the dependence of the performance effects of adaptive systems on system hardware, algorithms, and algorithm parameters.

The Effects of a Scheduled Driver Engagement Strategy in Automated Driving

This simulator study investigated a driver engagement (DE) strategy designed to keep the driver-in-the-loop during automated driving in the face of two different types of secondary tasks. The method, first reported by Carsten et al. (2012), involved driving in fully automated driving mode for 6 minutes followed by 1 minute of manual driving, after which this fixed schedule was repeated several times throughout the drive. This scheduled strategy was compared to a reference condition in which different participants experienced continuous automated driving without interruptions. For each condition, some participants watched a video and others listened to the radio. All drives ended in automated driving mode with a surprise forward collision (FC) hazard to which the participant had to manually intervene. Compared to video watchers, radio listeners responded faster, looked to the road scene more, and they were more often looking forward at FC event onset. The DE strategy had no effect on radio listeners. In contrast, video watchers responded to the hazard more quickly with the scheduled strategy than without it. However, there was no reliable statistical difference between DE conditions in percent-eye-glance-time looking to the forward road scene during automated driving or in the number of drivers looking forward at FC event onset. Thus, the scheduled driver engagement strategy appears beneficial in the face of the visual (video) secondary task, but the reasons for this advantage remain to be determined.

Simulator Study of Effects of Alternative Distraction Mitigation Strategies in Driver Workload Manager

This simulator study examined a workload manager developed by Delphi Electronics for the SAVE-IT program and the effects of several different workload mitigation strategies on driver response to a surprise forward collision hazard. The strategies included no in-vehicle task or distraction (baseline); task allowed; task interrupted; and task denied. Forty-eight test participants (24 males and 24 females) between 35 and 55 years of age were randomly assigned in groups of 12 (balanced for gender) to each of the four conditions. Each participant then drove in the Ford VIRTTEX moving-base driving simulator on simulated urban and rural roads and was asked to perform various in-vehicle tasks. During a requested in-vehicle information system task, a vehicle parked on the side of the road would suddenly enter the travel lane, and the drivers response was assessed. Braking response to this critical event indicated no significant differences in mean brake response time as a function of type of mitigation strategy or gender. However, variability in driver responses was significantly less in the task denied condition as compared with the other conditions, possibly because drivers were sensitized to an increased driving demand. Three of 12 test participants in the task interrupted condition showed relatively large brake reaction times attributable to long delays between initial foot motion and braking onset. This delay may indicate an additional delay associated with processing the task interruption and the forward collision warning event itself. Recommendations are provided for further research and for mitigation and driver alerting on the basis of a workload managers assessment of the driving situation.

Eye Glance and Head Turn Correspondence during Secondary Task Performance in Simulator Driving

This paper and its findings serve as a fundamental characterization of the interaction between the head and eye as a function of task demand and duration while driving. These findings have potential use in future vehicle active safety, and the design of driver assistance and/or HMI systems and features. Our findings show that head position and eye-point-of regard for certain task locations have dependence and may be suited to characterize task engagement while driving.

Driver response to unexpected automatic braking/haptic warning while backing

Two studies were conducted to examine driver behavior in response to an unexpected automatic braking/haptic event while backing. One study emulated a False Positive condition (no object present in reversing path at/after braking event onset). A second study involved a True Positive condition (obstacle present at/after braking event onset). Results of the False Positive study indicated that, at braking event onset, drivers either already had their foot on the brake, or quickly placed it there. Shortly thereafter many drivers began to exhibit exploratory behavior to test the state of their vehicle, e.g. release of brake pedal, squeezing the accelerator pedal, etc. Results of the True Positive study also indicated that drivers either already had their foot on the brake or quickly placed it there. However, a higher percentage of participants held the brake pressed. Their exploratory behavior to test the state of the vehicle involved brake pedal release only. No driver pushed the accelerator pedal when an obstacle was visible in the rear-camera display. Driver eye gaze and head movements were observed for both False Positive and True Positive scenarios. Results are presented and their implications are discussed.

A Measurement Systems Analysis of Total Shutter Open Time (TSOT) as a Distraction Metric for Visual-Manual Tasks

The Total Shutter Open Time (TSOT) metric was examined for estimating the visual-manual distraction potential of in-vehicle devices. A measurement systems analysis was carried out on TSOT using data on thirteen visual-manual tasks from the CAMP Driver Workload Metrics Project. TSOT showed low test-retest reliability but high repeatability when data were averaged across persons by task. TSOT predicted task completion time, lane keeping, speed variation, total glance time, and number of glances away from the road while driving. Tasks were classified into higher and lower workload categories based on literature, analytical modeling, and engineering judgment. TSOT showed a high percentage of statistically significant pairwise differences between higher vs. lower workload tasks. Different classification rules were also applied to TSOT. The best rule to classify tasks as higher or lower workload consistent with prior prediction was one in which a mean TSOT > 7.5 seconds implied the task was of higher workload. These results illustrate a general procedure to assess driver workload measures in general and the usefulness of TSOT in particular.

Off-Road Glance Behavior, Response Time to a Forward Collision Hazard, and Engagement Strategy Effects in Automated Driving

Blommer et al. (2015) reported on a simulator study that investigated a driver engagement (DE) strategy designed to keep the driver-in-the-loop during automated driving in the face of two different types of secondary tasks. The method, first reported by Carsten et al. (2012), involved driving in fully automated driving mode for 6 minutes followed by 1 minute of manual driving, after which this fixed schedule was repeated several times throughout the drive. This scheduled strategy was compared to a reference condition in which different participants experienced continuous automated driving without interruptions. For each condition, some participants watched a video and others listened to the radio. All drives ended in automated driving mode with a surprise forward collision (FC) hazard to which the participant had to manually intervene. Compared to video watchers, radio listeners responded faster, looked to the road scene more, and they were more often looking forward at FC event onset. The DE strategy had no effect on radio listeners. In contrast, video watchers responded to the hazard more quickly with the scheduled strategy than without it. However, there was no reliable statistical difference between DE conditions in percent-eye-glance-time looking to the forward road scene during automated driving or in the number of drivers looking forward at FC event onset. This paper presents additional analyses of off-road eye glance behavior and finds no relationship between how long people were looking away prior to receiving a Forward Collision Warning (FCW) and driver response time (RT). About 95% of all video watching drivers glanced back to the road within 20 sec regardless of the automated driving condition. Approximately 85% of glances away from the road in the scheduled mitigation condition were 7 sec or less.

MTS-DeepNet for lane change prediction.

Time series data are ubiquitous and are of importance in many application problems in engineering, science, medicine, economics and entertainment. Many real world pattern classification problems involve the processing and analysis of multiple variables in the temporal domain. These types of problems are referred to as Multivariate Time Series (MTS) problems. In many real-world applications, an MTS problem can involve a large number of signals, and require algorithms to select signals and extract temporal and spatial features from them. In this paper, we present an innovative convolutional neural network, MTS-DeepNet that is specially designed for MTS pattern classification. The system integrates signal and feature selections with MTS pattern classification in one learning framework. MTS-DeepNet is applied to a real-world problem, namely predicting driver lane departure based on drivers physiological signals. Our experimental results showed that, in comparison to a multi-layer neural network trained with the backpropagation algorithm, MTS-DeepNet gave better prediction accuracy.

An Exploratory Study of Driver Response to Reduced System Confidence Notifications in Automated Driving

An exploratory study in a moving-base simulator with 34 volunteers examined the effects of automated driving system reduced-confidence notifications on driver response. Reduced confidence in this context is defined as a system performance level that is lower than optimal, yet not so low as to disable the system. The threshold for such a level is essentially a policy decision based on perceived costs and benefits of keeping the automation engaged versus disengaging it. So, a reduced-confidence notification was designed to simply indicate that the driver should be especially vigilant to the increased potential to have to take manual control of the vehicle. An actual manual takeover request from the automated system would require the driver to take manual control of the vehicle. To investigate this, a 32-min highway driving scenario was constructed for this exploratory investigation that included 16 potential loss-of-lane tracking segments within this time frame. Three factors at two levels (low/high) each were crossed and presented in a between-subjects experimental design. One factor was system confidence/competence match, which refers to the ratio of the number of actual automation system failures over the number of reduced-confidence notifications. A second factor was the consequence of automation failure, i.e., drifting toward a busy traffic lane or drifting toward a wide shoulder. The third factor was detectability, or being able to see versus not see why the automated driving system began to drift out-of-lane. Data analyses indicated that automated driving with reduced-confidence notifications was judged to be both useful and acceptable regardless of the manipulated factors. However, the lowest incidence of preemption (i.e., manual takeovers prior to the start of a lane drift) was associated with low automation competence/confidence match and low detectability, regardless of consequences. It was also associated with responding later by some drivers. On the other hand, when the reduced-confidence notifications were perfectly matched with the need to manually takeover, there were fewer instances of delayed driver responses. Results are discussed in terms of “cry wolf” effects.