Timothy L. Brown

Collision Warning Timing, Driver Distraction, and Driver Response to Imminent Rear-End Collisions in a High-Fidelity Driving Simulator:

Rear-end collisions account for almost 30% of automotive crashes. Rear-end collision avoidance systems (RECASs) may offer a promising approach to help drivers avoid these crashes. Two experiments performed using a high-fidelity motion-based driving simulator examined driver responses to evaluate the efficacy of a RECAS. The first experiment showed that early warnings helped distracted drivers react more quickly---and thereby avoid more collisions---than did late warnings or no warnings. Compared with the no-warning condition, an early RECAS warning reduced the number of collisions by 80.7%. Assuming collision severity is proportional to kinetic energy, the early warning reduced collision severity by 96.5%. In contrast, the late warning reduced collisions by 50.0 % and the corresponding severity by 87.5%. The second experiment showed that RECAS benefits even undistracted drivers. Analysis of the braking process showed that warnings provide a potential safety benefit by reducing the time required for drivers to release the accelerator. Warnings do not, however, speed application of the brake, increase maximum deceleration, or affect mean deceleration. These results provide the basis for a computational model of driver performance that was used to extrapolate the findings and identify the most promising parameter settings. Potential applications of these results include methods for evaluating collision warning systems, algorithm design guidance, and driver performance model input.

Human performance models and rear-end collision avoidance algorithms

Collision warning systems offer a promising approach to mitigate rear-end collisions, but substantial uncertainty exists regarding the joint performance of the driver and the collision warning algorithms. A simple deterministic model of driver performance was used to examine kinematics-based and perceptual-based rear-end collision avoidance algorithms over a range of collision situations, algorithm parameters, and assumptions regarding driver performance. The results show that the assumptions concerning driver reaction times have important consequences for algorithm performance, with underestimates dramatically undermining the safety benefit of the warning. Additionally, under some circumstances, when drivers rely on the warning algorithms, larger headways can result in more severe collisions. This reflects the nonlinear interaction among the collision situation, the algorithm, and driver response that should not be attributed to the complexities of driver behavior but to the kinematics of the situation. Comparisons made with experimental data demonstrate that a simple human performance model can capture important elements of system performance and complement expensive human-in-the-loop experiments. Actual or potential applications of this research include selection of an appropriate algorithm, more accurate specification of algorithm parameters, and guidance for future experiments.

Effects of Adaptive Cruise Control and Alert Modality on Driver Performance

This article reports on a study that assessed the ability of automobile drivers to make the transition from adaptive cruise control (ACC) to manual control when warned with alerts of different modalities. The authors compared how drivers maintain headway distance in conditions with and without ACC in mild, moderate, and severe braking situations. The different modalities tested include visual, auditory, seat vibration, brake pulse, and a combination of these methods. The two scenarios studied were a braking lead vehicle and an abrupt lane change of a lead vehicle that reveals a slow-moving vehicle. The study included sixty people aged 30 to 50 years, split evenly by gender and used to using cruise control (defined as at least twice per month). After a demographic questionnaire and pre-drive instruction, participants drove a 6-minute practice drive, followed by a 35-minute experimental drive. The results showed that ACC helped drivers maintain a larger safety margin, as measured by the minimum time-to-collision (TTC). The authors hypothesize that this larger safety margin may have important indirect benefits, affecting other drivers and the overall traffic flow rather than the likelihood of a crash for the driver using the ACC. The various alert modalities performed similarly when considered independently. There was a slightly greater minimum TTC associated with the brake pulse in moderately severe situations. Readers are referred to the full report at www.ntis.gov (access number PB2009-102474). Keywords: Driver distraction;

Cannabis effects on driving lateral control with and without alcohol

BACKGROUND Effects of cannabis, the most commonly encountered non-alcohol drug in driving under the influence cases, are heavily debated. We aim to determine how blood Δ(9)-tetrahydrocannabinol (THC) concentrations relate to driving impairment, with and without alcohol. METHODS Current occasional (≥1×/last 3 months, ≤3days/week) cannabis smokers drank placebo or low-dose alcohol, and inhaled 500mg placebo, low (2.9%)-THC, or high (6.7%)-THC vaporized cannabis over 10min ad libitum in separate sessions (within-subject design, 6 conditions). Participants drove (National Advanced Driving Simulator, University of Iowa) simulated drives (∼0.8h duration). Blood, oral fluid (OF), and breath alcohol samples were collected before (0.17h, 0.42h) and after (1.4h, 2.3h) driving that occurred 0.5-1.3h after inhalation. We evaluated standard deviations of lateral position (lane weave, SDLP) and steering angle, lane departures/min, and maximum lateral acceleration. RESULTS In N=18 completers (13 men, ages 21-37years), cannabis and alcohol increased SDLP. Blood THC concentrations of 8.2 and 13.1μg/L during driving increased SDLP similar to 0.05 and 0.08g/210L breath alcohol concentrations, the most common legal alcohol limits. Cannabis-alcohol SDLP effects were additive rather than synergistic, with 5μg/L THC+0.05g/210L alcohol showing similar SDLP to 0.08g/210L alcohol alone. Only alcohol increased lateral acceleration and the less-sensitive lane departures/min parameters. OF effectively documented cannabis exposure, although with greater THC concentration variability than paired blood samples. CONCLUSIONS SDLP was a sensitive cannabis-related lateral control impairment measure. During drive blood THC ≥8.2μg/L increased SDLP similar to notably-impairing alcohol concentrations. Despite OFs screening value, OF variability poses challenges in concentration-based effects interpretation.

Effect of warning timing on collision avoidance behavior in a stationary lead vehicle scenario

Warning timing and how drivers with and without forward collision warning (FCW) systems react when distracted at the moment a stationary vehicle is revealed directly ahead were investigated. The study was conducted using the Iowa Driving Simulator (IDS). The IDS was equipped with an FCW system that provided auditory warnings based on two warning criteria. A total of 30 subjects were split across three conditions—a baseline of 10 subjects (no warning display), and two warning conditions (early and late) with 10 subjects each. The two warning conditions differed by the duration of an a priori driver reaction component (1.5 and 1.0 s) in the warning algorithm. Drivers’ collision avoidance performance in the two warning conditions was compared with that in the baseline condition. Results indicated that the early warning condition showed significantly shorter accelerator release reaction times, fewer crashes, and less severe crashes than both the baseline condition and the late warning condition. The results indicate that the timing of a warning is important in the design of collision warning systems.

Comparison of Driver Braking Responses in a High-Fidelity Simulator and on a Test Track

The braking responses of drivers in the Iowa Driving Simulator (IDS) were compared with those of drivers on a test track. The braking profile of drivers was compared in last-minute braking situations in which drivers were instructed to brake “normally” or “hard.” Although the motion and visual cues in the IDS are imperfect, the data agree in many respects. The general pattern of results is similar, with the initial speed and lead vehicle deceleration affecting drivers on the test track and in the simulator in a similar way. In several experimental conditions, the similarity of the responses went beyond the general pattern of response. The mean values were almost identical in several instances, and the values were frequently well within the confidence intervals. Although the simulator and test track drivers performed similarly, differences are apparent from the onset of braking, through the braking process, and in the outcome of the braking event. The instructions concerning normal and hard braking had little influence on the behavior of drivers in the simulator. Contributors to these differences include the limited visual and vestibular cues in the simulator and the extended practice on the test track.

Controlled Cannabis Vaporizer Administration: Blood and Plasma Cannabinoids with and without Alcohol

BACKGROUND Increased medical and legal cannabis intake is accompanied by greater use of cannabis vaporization and more cases of driving under the influence of cannabis. Although simultaneous Δ(9)-tetrahydrocannabinol (THC) and alcohol use is frequent, potential pharmacokinetic interactions are poorly understood. Here we studied blood and plasma vaporized cannabinoid disposition, with and without simultaneous oral low-dose alcohol. METHODS Thirty-two adult cannabis smokers (≥1 time/3 months, ≤3 days/week) drank placebo or low-dose alcohol (target approximately 0.065% peak breath-alcohol concentration) 10 min before inhaling 500 mg placebo, low-dose (2.9%) THC, or high-dose (6.7%) THC vaporized cannabis (6 within-individual alcohol-cannabis combinations). Blood and plasma were obtained before and up to 8.3 h after ingestion. RESULTS Nineteen participants completed all sessions. Median (range) maximum blood concentrations (Cmax) for low and high THC doses (no alcohol) were 32.7 (11.4-66.2) and 42.2 (15.2-137) μg/L THC, respectively, and 2.8 (0-9.1) and 5.0 (0-14.2) μg/L 11-OH-THC. With alcohol, low and high dose Cmax values were 35.3 (13.0-71.4) and 67.5 (18.1-210) μg/L THC and 3.7 (1.4-6.0) and 6.0 (0-23.3) μg/L 11-OH-THC, significantly higher than without alcohol. With a THC detection cutoff of ≥1 μg/L, ≥16.7% of participants remained positive 8.3 h postdose, whereas ≤21.1% were positive by 2.3 h with a cutoff of ≥5 μg/L. CONCLUSIONS Vaporization is an effective THC delivery route. The significantly higher blood THC and 11-OH-THC Cmax values with alcohol possibly explain increased performance impairment observed from cannabis-alcohol combinations. Chosen driving-related THC cutoffs should be considered carefully to best reflect performance impairment windows. Our results will help facilitate forensic interpretation and inform the debate on drugged driving legislation.

Use patterns among early adopters of adaptive cruise control

Objective: The objective of this study was to investigate use patterns among early adopters of adaptive cruise control (ACC). Background: Extended use of ACC may influence a driver’s behavior in the long term, which can have unintended safety consequences. Method: The authors examined the use of a motion-based simulator by 24 participants (15 males and 9 females). Cluster analysis was performed on drivers’ use of ACC and was based on their gap settings, speed settings, number of warnings issued, and ACC disengaged. The data were then examined on the basis of driving performance measures and drivers’ subjective responses to trust in ACC, understanding of system operations, and driving styles. Driving performance measures included minimum time headway, adjusted minimum time to collision, and drivers’ reaction time to critical events. Results: Three groups of drivers were observed on the basis of risky behavior, moderately risky behavior, and conservative behavior. Drivers in the conservative group stayed farther behind the lead vehicle than did drivers in the other two groups. Risky drivers responded later to critical events and had more ACC warnings issued. Conclusion: Safety consequences with ACC may be more prevalent in some driver groups than others. The findings suggest that these safety implications are related to trust in automation, driving styles, understanding of system operations, and personalities. Application: Potential applications of this research include enhanced design for next-generation ACC systems and countermeasures to improve safe driving with ACC.

An empirical study of the effectiveness of electronic stability control system in reducing loss of vehicle control.

A significant percentage of fatal vehicle crashes involve loss of control (LOC). Electronic stability control (ESC) is an active safety system that detects impending LOC and activates counter-measures that help the driver maintain or re-gain control. To assess the effectiveness of ESC in preventing LOC, an empirical study was conducted on a high-fidelity driving simulator. The ESC systems for two vehicles were incorporated into the simulators dynamics code which was calibrated to ensure engineering validation. The study utilized three scenarios designed to recreate typical LOC situations, and was designed to assess the effects of ESC presence, vehicle type, scenario, age and gender. A total of 120 research participants completed the study. Results showed a statistically significant reduction in LOC with ESC compared to without ESC (F=52.72, p<0.0001). The study findings of 5% LOC with ESC and 30% without ESC match several epidemiological studies that have analyzed ESC effectiveness on real-world crashes, providing strong support to the use of driving simulation for studying driver behavior. Study conclusions suggest that wide-spread utilization of ESC is likely to reduce traffic fatalities.

Steering in a random forest: ensemble learning for detecting drowsiness-related lane departures

Objective: The aim of this study was to design and evaluate an algorithm for detecting drowsiness-related lane departures by applying a random forest classifier to steering wheel angle data. Background: Although algorithms exist to detect and mitigate driver drowsiness, the high rate of false alarms and missed detection of drowsiness represent persistent challenges. Current algorithms use a variety of data sources, definitions of drowsiness, and machine learning approaches to detect drowsiness. Method: We develop a new approach for detecting drowsiness-related lane departures using steering wheel angle data that employ an ensemble definition of drowsiness and a random forest algorithm. Data collected from 72 participants driving the National Advanced Driving Simulator are used to train and evaluate the model. The model’s performance was assessed relative to a commonly used algorithm, percentage eye closure (PERCLOS). Results: The random forest steering algorithm had a higher classification accuracy and area under the receiver operating characteristic curve than PERCLOS and had comparable positive predictive value. The algorithm succeeds at identifying two key scenarios associated with the drowsiness detection task. These two scenarios consist of instances when drivers depart their lane because they fail to modulate their steering behavior according to the demands of the simulated road and instances when drivers correctly modulate their steering behavior according to the demands of the road. Conclusion: The random forest steering algorithm is a promising approach to detect driver drowsiness. The algorithm’s ties to consequences of drowsy driving suggest that it can be easily paired with mitigation systems.