Improvement of driver active interventions during automated driving by displaying trajectory pointers—A driving simulator study

Abstract

Objective: The handover of vehicle control from automated to manual operation is a critical aspect of interaction between drivers and automated driving systems (ADS). In some cases, it is possible that the ADS may fail to detect an object. In this event, the driver must be aware of the situation and resume control of the vehicle without assistance from the system. Consequently, the driver must fulfill the following 2 main roles while driving: (1) monitor the vehicle trajectory and surrounding traffic environment and (2) actively take over vehicle control if the driver identifies a potential issue along the trajectory. An effective human–machine interface (HMI) is required that enables the driver to fulfill these roles. This article proposes an HMI that constantly indicates the future position of the vehicle.

Methods: This research used the Toyota Dynamic Driving Simulator to evaluate the effect of the proposed HMI and compares the proposed HMI with an HMI that notifies the driver when the vehicle trajectory changes. A total of 48 test subjects were divided into 2 groups of 24: One group used the HMI that constantly indicated the future position of the vehicle and the other group used the HMI that provided information when the vehicle trajectory changed.

The following instructions were given to the test subjects: (1) to not hold the steering wheel and to allow the vehicle to drive itself, (2) to constantly monitor the surrounding traffic environment because the functions of the ADS are limited, and (3) to take over driving if necessary.

The driving simulator experiments were composed of an initial 10-min acclimatization period and a 10-min evaluation period. Approximately 10?min after the start of the evaluation period, a scenario occurred in which the ADS failed to detect an object on the vehicle trajectory, potentially resulting in a collision if the driver did not actively take over control and manually avoid the object.

Results: The collision avoidance rate of the HMI that constantly indicated the future position of the vehicle was higher than that of the HMI that notified the driver of trajectory changes, χ² = 6.38, P < .05. The steering wheel hands-on and steering override timings were also faster with the proposed HMI (t test; P < .05).

Conclusions: This research confirmed that constantly indicating the position of the vehicle several seconds in the future facilitates active driver intervention when an ADS is in operation.     

Characteristics of rear-end crashes involving passenger vehicles with automatic emergency braking

Abstract

Objectives: Automatic emergency braking (AEB) is a proven effective countermeasure for preventing front-to-rear crashes, but it has not yet fully lived up to its estimated potential. This study identified the types of rear-end crashes in which striking vehicles with AEB are overrepresented to determine whether the system is more effective in some situations than in others, so that additional opportunities for increasing AEB effectiveness might be explored.

Methods: Rear-end crash involvements were extracted from 23?U.S. states during 2009–2016 for striking passenger vehicles with and without AEB among models where the system was optional. Logistic regression was used to examine the odds that rear-end crashes with various characteristics involved a striking vehicle with AEB, controlling for driver and vehicle features.

Results: Striking vehicles were significantly more likely to have AEB in crashes where the striking vehicle was turning relative to when it was moving straight (odds ratio [OR]?=?2.35; 95% confidence interval [CI], 1.76, 3.13); when the struck vehicle was turning (OR = 1.66; 95% CI, 1.25, 2.21) or changing lanes (OR = 2.05; 95% CI, 1.13, 3.72) relative to when it was slowing or stopped; when the struck vehicle was not a passenger vehicle or was a special use vehicle relative to a car (OR = 1.61; 95% CI, 1.01, 2.55); on snowy or icy roads relative to dry roads (OR = 1.83; 95% CI, 1.16, 2.86); or on roads with speed limits of 70+ mph relative to those with 40 to 45?mph speed limits (OR = 1.49; 95% CI, 1.10, 2.03). Overall, 25.3% of crashes where the striking vehicle had AEB had at least one of these overrepresented characteristics, compared with 15.9% of strikes by vehicles without AEB.

Conclusions: The typical rear-end crash occurs when 2 passenger vehicles are proceeding in line, on a dry road, and at lower speeds. Because atypical crash circumstances are overrepresented among rear-end crashes by striking vehicles with AEB, it appears that the system is doing a better job of preventing the more typical crash scenario. Consumer information testing programs of AEB use a test configuration that models the typical rear-end crash type. Testing programs promoting good AEB performance in crash circumstances where vehicles with AEB are overrepresented could guide future development of AEB systems that perform well in these additional rear-end collision scenarios.     

Effects of lead time of verbal collision warning messages on driving behavior in connected vehicle settings

IntroductionUnder the connected vehicle environment, vehicles will be able to exchange traffic information with roadway infrastructure and other vehicles. With such information, collision warning systems (CWSs) will be able to warn drivers with potentially hazardous situations within or out of sight and reduce collision accidents. The lead time of warning messages is a crucial factor in determining the effectiveness of CWSs in the prevention of traffic accidents. Accordingly, it is necessary to understand the effects of lead time on driving behaviors and explore the optimal lead time in various collision scenarios.MethodsThe present driving simulator experiment studied the effects of controlled lead time at 16 levels (predetermined time headway from the subject vehicle to the collision location when the warning message broadcasted to a driver) on driving behaviors in various collision scenarios.ResultsMaximum effectiveness of warning messages was achieved when the controlled lead time was within the range of 5 s to 8 s. Specifically, the controlled lead time ranging from 4 s to 8 s led to the optimal safety benefit; and the controlled lead time ranging from 5 s to 8 s led to more gradual braking and shorter reaction time. Furthermore, a trapezoidal distribution of warning effectiveness was found by building a statistic model using curve estimation considering lead time, lifetime driving experience, and driving speed.ConclusionsThe results indicated that the controlled lead time significantly affected driver performance.Practical applicationsThe findings have implications for the design of collision warning systems.     

Safety assessment of automated vehicles: how to determine whether we have collected enough field data?

Abstract

Objective: The amount of collected field data from naturalistic driving studies is quickly increasing. The data are used for, among others, developing automated driving technologies (such as crash avoidance systems), studying driver interaction with such technologies, and gaining insights into the variety of scenarios in real-world traffic. Because data collection is time consuming and requires high investments and resources, questions like “Do we have enough data?,” “How much more information can we gain when obtaining more data?,” and “How far are we from obtaining completeness?” are highly relevant. In fact, deducing safety claims based on collected data—for example, through testing scenarios based on collected data—requires knowledge about the degree of completeness of the data used. We propose a method for quantifying the completeness of the so-called activities in a data set. This enables us to partly answer the aforementioned questions.

Method: In this article, the (traffic) data are interpreted as a sequence of different so-called scenarios that can be grouped into a finite set of scenario classes. The building blocks of scenarios are the activities. For every activity, there exists a parameterization that encodes all information in the data of each recorded activity. For each type of activity, we estimate a probability density function (pdf) of the associated parameters. Our proposed method quantifies the degree of completeness of a data set using the estimated pdfs.

Results: To illustrate the proposed method, 2 different case studies are presented. First, a case study with an artificial data set, of which the underlying pdfs are known, is carried out to illustrate that the proposed method correctly quantifies the completeness of the activities. Next, a case study with real-world data is performed to quantify the degree of completeness of the acquired data for which the true pdfs are unknown.

Conclusion: The presented case studies illustrate that the proposed method is able to quantify the degree of completeness of a small set of field data and can be used to deduce whether sufficient data have been collected for the purpose of the field study. Future work will focus on applying the proposed method to larger data sets. The proposed method will be used to evaluate the level of completeness of the data collection on Singaporean roads, aimed at defining relevant test cases for the autonomous vehicle road approval procedure that is being developed in Singapore.     

Creating Pedestrian Crash Scenarios in a Driving Simulator Environment

Objective: In 2012 in the United States, pedestrian injuries accounted for 3.3% of all traffic injuries but, disproportionately, pedestrian fatalities accounted for roughly 14% of traffic-related deaths (NHTSA 2014 NHTSA. Traffic Safety Facts 2012 Pedestrians. Washington, DC: Author; 2014. DOT HS 811 888. [Google Scholar]). In many other countries, pedestrians make up more than 50% of those injured and killed in crashes. This research project examined driver response to crash-imminent situations involving pedestrians in a high-fidelity, full-motion driving simulator. This article presents a scenario development method and discusses experimental design and control issues in conducting pedestrian crash research in a simulation environment. Driving simulators offer a safe environment in which to test driver response and offer the advantage of having virtual pedestrian models that move realistically, unlike test track studies, which by nature must use pedestrian dummies on some moving track.

Methods: An analysis of pedestrian crash trajectories, speeds, roadside features, and pedestrian behavior was used to create 18 unique crash scenarios representative of the most frequent and most costly crash types. For the study reported here, we only considered scenarios where the car is traveling straight because these represent the majority of fatalities. We manipulated driver expectation of a pedestrian both by presenting intersection and mid-block crossing as well as by using features in the scene to direct the driver's visual attention toward or away from the crossing pedestrian. Three visual environments for the scenarios were used to provide a variety of roadside environments and speed: a 20–30 mph residential area, a 55 mph rural undivided highway, and a 40 mph urban area.

Results: Many variables of crash situations were considered in selecting and developing the scenarios, including vehicle and pedestrian movements; roadway and roadside features; environmental conditions; and characteristics of the pedestrian, driver, and vehicle. The driving simulator scenarios were subjected to iterative testing to adjust time to arrival triggers for the pedestrian actions. This article discusses the rationale behind creating the simulator scenarios and some of the procedural considerations for conducting this type of research.

Conclusions: Crash analyses can be used to construct test scenarios for driver behavior evaluations using driving simulators. By considering trajectories, roadway, and environmental conditions of real-world crashes, representative virtual scenarios can serve as safe test beds for advanced driver assistance systems. The results of such research can be used to inform pedestrian crash avoidance/mitigation systems by identifying driver error, driver response time, and driver response choice (i.e., steering vs. braking).     

Fast Calculating Surrogate Models for Leg and Head Impact in Vehicle–Pedestrian Collision Simulations

Objective: In previous research, a tool chain to simulate vehicle–pedestrian accidents from ordinary driving state to in-crash has been developed. This tool chain allows for injury criteria-based, vehicle-specific (geometry, stiffness, active safety systems, etc.) assessments. Due to the complex nature of the included finite element analysis (FEA) models, calculation times are very high. This is a major drawback for using FEA models in large-scale effectiveness assessment studies. Therefore, fast calculating surrogate models to approximate the relevant injury criteria as a function of pedestrian vehicle collision constellations have to be developed.

Method: The development of surrogate models for head and leg injury criteria to overcome the problem of long calculation times while preserving high detail level of results for effectiveness analysis is shown in this article. These surrogate models are then used in the tool chain as time-efficient replacements for the FEA model to approximate the injury criteria values. The method consists of the following steps: Selection of suitable training data sets out of a large number of given collision constellations, detailed FEA calculations with the training data sets as input, and training of the surrogate models with the FEA model's input and output values.

Results: A separate surrogate model was created for each injury criterion, consisting of a response surface that maps the input parameters (i.e., leg impactor position and velocity) to the output value. In addition, a performance test comparing surrogate model predictions of additional collision constellations to the results of respective FEA calculations was carried out. The developed method allows for prediction of injury criteria based on impact constellation for a given vehicle. Because the surrogate models are specific to a certain vehicle, training has to be redone for a new vehicle. Still, there is a large benefit regarding calculation time when doing large-scale studies.

Conclusion: The method can be used in prospective effectiveness assessment studies of new vehicle safety features and takes into account specific local features of a vehicle (geometry, stiffness, etc.) as well as external parameters (location and velocity of pedestrian impact). Furthermore, it can be easily extended to other injury criteria or accident scenarios; for example, cyclist accidents.     

Field effectiveness evaluation of advanced driver assistance systems

Abstract

Objective: Advanced driver assistance systems (ADAS) are a class of vehicle technologies designed to increase safety by providing drivers with timely warnings and autonomously intervening to avoid hazardous situations. Though laboratory testing suggests that ADAS technologies will greatly impact crash involvement rates, real-world evidence that characterizes their effectiveness is still limited. This study evaluates and quantifies the association of ADAS technologies with the likelihood of a moderate or severe crash for new-model BMWs in the United States.

Methods: Vehicle ADAS option information for the cohort of model year 2014 and later BMW passenger vehicles sold after January 1, 2014 (n?=?1,063,503), was coded using VIN-identified options data. ADAS technologies of interest include frontal collision warning with autonomous emergency braking, lane departure warning, and blind spot detection. BMW Automated Crash Notification system data (from January 2014 to November 2017) were merged with vehicle data by VIN to identify crashed vehicles (n?=?15,507), including date, crash severity (delta V), and area of impact. Using Cox proportional hazards regression modeling, the study calculates the adjusted hazard ratio for crashing among BMW passenger vehicles with versus without ADAS technologies. The adjusted percentage reduction in moderate and severe crashes associated with ADAS is interpreted as one minus the hazard ratio.

Results: Vehicles equipped with both autonomous emergency braking and lane departure warning were 23% less likely to crash than those not equipped (hazard ratio [HR]?=?0.77; 95% confidence interval [CI], 0.73–0.81), controlling for model year, vehicle size and body type. Autonomous emergency braking and lane departure warning generally occur together, making it difficult to tease apart their individual effects. Blind spot detection was associated with a 14% reduction in crashes after controlling for the presence of autonomous emergency braking and lane departure warning (HR =0.86; 95% CI, 0.744–0.99). Differences were observed by vehicle type and crash type. The combined effect of autonomous emergency braking and lane departure warning was greater in newer model vehicles: Equipped vehicles were 13% less likely to crash (HR =0.87; 95% CI, 0.79–0.95) among 2014 model year vehicles versus 34% less likely to crash (HR =0.66; 95% CI, 0.57–0.77) among 2017 model year vehicles.

Conclusion: This robust cohort study contributes to the growing evidence on the effectiveness of ADAS technologies.     

How do drivers avoid collisions? A driving simulator-based study

Introduction: Drivers' collision avoidance performance in an impending collision situation plays a decisive role for safety outcomes. This study explored drivers' collision avoidance performances in three typical collision scenarios that were right-angle collision, head-on collision, and collision with pedestrian. Method: A high-fidelity driving simulator was used to design the scenarios and conduct the experiment. 45 participants took part in the simulator experiment. Drivers' longitudinal/lateral collision avoidance performances and collision result were recorded. Results: Experimental results showed that brake only was the most common response among the three collision scenarios, followed by brake combining swerve in head-on and pedestrian collision scenarios. In right-angle collision scenario with TTC (time to collision) largest among three scenarios, no driver swerved, and meanwhile drivers who showed slow brake reaction tended to compensate the collision risk by taking a larger maximum deceleration rate within a shorter time. Swerve-toward-conflict was a prevalent phenomenon in head-on and pedestrian collision scenarios and significantly associated with collision risk. Drivers that swerved toward the conflict object had a shorter swerve reaction time than drivers that swerved away from conflict. Conclusions: Long brake reaction time and wrong swerve direction were the main factors leading to a high collision likelihood. The swerve-toward-conflict maneuver caused a delay in brake action and degraded subsequent braking performances. The prevalent phenomenon indicated that drivers tended to use an intuitive (heuristic) way to make decisions in critical traffic situations. Practical applications: The study generated a better understanding of collision development and shed lights on the design of future advanced collision avoidance systems for semi-automated vehicles. Manufactures should also engage more efforts in developing active steering assistance systems to assist drivers in collision avoidance.     

Multi-agent traffic simulations to estimate the impact of automated technologies on safety

Abstract

Objective: The objective of this article was to develop a multi-agent traffic simulation methodology to estimate the potential road safety improvements of automated vehicle technologies.

Methods: We developed a computer program that merges road infrastructure data with a large number of vehicles, drivers, and pedestrians. Human errors are induced by modeling inattention, aimless driving, insufficient safety confirmation, misjudgment, and inadequate operation. The program was applied to simulate traffic in a prescribed area in Tsukuba city. First, a 100% manual driving scenario was set to simulate traffic for a total preset vehicle travel distance. The crashes from this simulation were compared with real-world crash data from the prescribed area from 2012 to 2017. Thereafter, 4 additional scenarios of increasing levels of automation penetration (including combinations of automated emergency braking [AEB], lane departure warning [LDW], and SAE Level 4 functions) were implemented to estimate their impact on safety.

Results: Under manual driving, the system simulated a total of 859 crashes including single-car lane departure, car-to-car, and car-to-pedestrian crashes. These crashes tended to occur in locations similar to real-world crashes. The number of crashes predicted decreased to 156 cases with increasing level of automation. All of the technologies considered contributed to the decrease in crashes. Crash reductions attributable to AEB and LDW in the simulations were comparable to those reported in recent field studies. For the highest levels of automation, no assessment data were available and hence the results should be carefully treated. Further, in modeling automated functions, potentially negative aspects such as sensing failure or human overreliance were not incorporated.

Conclusions: We developed a multi-agent traffic simulation methodology to estimate the effect of different automated vehicle technologies on safety. The crash locations resulting from simulations of manual driving within a limited area in Japan were preliminary assessed by comparison with real-world crash data collected in the same area. Increasing penetration levels of AEB and LDW led to a large reduction in both the frequency and severity of rear-end crashes, followed by car-to-car head-on crashes and single-vehicle lane departure crashes. Preliminary estimations of the potential safety improvements that may be achieved with highly automated driving technologies were also obtained.     

Traffic accident reconstruction and an approach for prediction of fault rates using artificial neural networks: A case study in Turkey

Objective: Currently, in Turkey, fault rates in traffic accidents are determined according to the initiative of accident experts (no speed analyses of vehicles just considering accident type) and there are no specific quantitative instructions on fault rates related to procession of accidents which just represents the type of collision (side impact, head to head, rear end, etc.) in No. 2918 Turkish Highway Traffic Act (THTA 1983). The aim of this study is to introduce a scientific and systematic approach for determination of fault rates in most frequent property damage–only (PDO) traffic accidents in Turkey.

Methods: In this study, data (police reports, skid marks, deformation, crush depth, etc.) collected from the most frequent and controversial accident types (4 sample vehicle–vehicle scenarios) that consist of PDO were inserted into a reconstruction software called vCrash. Sample real-world scenarios were simulated on the software to generate different vehicle deformations that also correspond to energy-equivalent speed data just before the crash. These values were used to train a multilayer feedforward artificial neural network (MFANN), function fitting neural network (FITNET, a specialized version of MFANN), and generalized regression neural network (GRNN) models within 10-fold cross-validation to predict fault rates without using software. The performance of the artificial neural network (ANN) prediction models was evaluated using mean square error (MSE) and multiple correlation coefficient (R).

Results: It was shown that the MFANN model performed better for predicting fault rates (i.e., lower MSE and higher R) than FITNET and GRNN models for accident scenarios 1, 2, and 3, whereas FITNET performed the best for scenario 4. The FITNET model showed the second best results for prediction for the first 3 scenarios. Because there is no training phase in GRNN, the GRNN model produced results much faster than MFANN and FITNET models. However, the GRNN model had the worst prediction results. The R values for prediction of fault rates were close to 1 for all folds and scenarios.

Conclusions: This study focuses on exhibiting new aspects and scientific approaches for determining fault rates of involvement in most frequent PDO accidents occurring in Turkey by discussing some deficiencies in THTA and without regard to initiative and/or experience of experts. This study yields judicious decisions to be made especially on forensic investigations and events involving insurance companies. Referring to this approach, injury/fatal and/or pedestrian-related accidents may be analyzed as future work by developing new scientific models.     

Decomposition analysis of the effects of vehicle safety technologies on the motor vehicle collision–related mortality rate from 1994 to 2015

Abstract

Objective: Though the mortality rate for motor vehicle collisions (MVCs) has been decreasing since the 1960s with the advent of the first federal seat belt laws in 1968, MVC remains a leading cause of death for individuals aged 1 to 44 years. The purpose of this study is to examine the effects of frontal (FABs) and side airbags (SABs) and electronic stability control (ESC) on the components of the MVC mortality rate.

Methods: The MVC mortality rate from 1994 to 2015 was separated into its components of exposure of vehicles, exposure of travel, collision density, injury incidence, and case fatality rate. Year was categorized on the availability of safety technology in vehicles: 1994–1997 (first-generation FABs mandated), 1998–2001 (sled-certified, second-generation FABs mandated), 2002–2006 (increasing prevalence of SABs and ESC), 2007–2011 (advanced airbags mandated), and 2012–2015 (ESC mandated, SAB in over 90% of vehicles, introduction of advanced safety systems). Relative contributions (RCs) of the components to changes in the MVC-related mortality rate were calculated as the absolute value of the component’s beta coefficient divided by the sum of the absolute values of all components’ beta coefficients. Negative binomial regression–estimated rate ratios (RRs) for the changes in the rate of each component by year category compared to the prior year category.

Results: Significant decreases in the MVC mortality rate were observed for 2007–2011 and 2012–2015. The decrease in 2007–2011 was due in most part to an 18% decrease in the injury incidence (RR?=?0.82, P?<?.0001, RC?=?63%), though there was a noted contribution by the decrease in vehicle miles traveled (RR?=?0.95, P?<?.0001, RC?=?15%). The continued decrease in mortality in 2012–2015 was due is most part to the 10% decreased case fatality rate (RR?=?0.90, P?<?.0001, RC?=?66%) because there was no significant change in the vehicle miles traveled and injury incidence.

Conclusions: The results of this study highlight the effects of vehicle safety technologies on the MVC-related mortality rate and can help direct prevention efforts. Through the study period, there was no meaningful contribution to decreases in the MVC-related mortality rate due to components related to exposure (i.e., vehicles per population and the rate of vehicle miles traveled), suggesting that prevention efforts at decreasing exposure prevalence would have little effect on the MVC-related mortality rate. Instead, prevention efforts should continue to focus on event-phase methods to decrease injury occurrence and mitigate injury severity during the collision.     

Influence of the feedback links of connected and automated vehicle on rear-end collision risks with vehicle-to-vehicle communication

Objective: Connected and automated vehicles (CAV) can monitor multiple vehicles ahead via vehicle-to-vehicle (V2V) communication. Although feedback information from more vehicles ahead may be more helpful for anticipations, it also makes control more complex and increases the probability of data packet loss. Then it needs an appropriate number of CAV feedback links, and the maximum number may be not suitable. Therefore, this article focuses on the influence of CAV feedback links on rear-end collision risks.

Methods: To deal with this, stability analysis of a CAV car-following model was conducted to obtain the designs of CAV feedback gains for maintaining stable CAV flow. Simulation experiments were performed to describe a traffic accident on freeway, using car-following models of manually driven vehicles (MDVs) and CAV under different CAV penetration rates. Four scenarios are considered in simulation experiments; that is, the CAV monitors 1, 2, 3, and 4 preceding vehicles, respectively. Based on the simulation experiments, surrogate safety indicators, time-exposed time-to-collision (TET), and time-integrated time-to-collision (TIT) are used to evaluate risks of rear-end collisions.

Results: Results indicated that CAV helped to decrease the collision risks, especially the more serious collision risks with smaller threshold values of time-to-collision (TTC). In addition, the reductions in collision risks are more obvious when CAV feedback changes from one link to 2 links. In addition, reducing amplitudes are not significant if the CAV feedback is extended from 2 links to 3 or 4 links.

Conclusions: Two links of CAV feedback are appropriate when control complexity is a priority, whereas 3 links is the better choice when reductions in collisions are a priority. The findings of this study provide helpful reference for CAV control and design before larger-scale implementation in real vehicles.     

Analysis of effects of driver’s evasive action time on rear-end collision risk using a driving simulator

Introduction: Driver’s evasive action is closely associated with collision risk in a critical traffic event. To quantify collision risk, surrogate safety measures (SSMs) have been estimated using vehicle trajectories. However, vehicle trajectories cannot clearly capture presence and time of driver’s evasive action. Thus, this study determines the driver’s evasive action based on his/her use of accelerator and brake pedals, and analyzes the effects of the driver’s evasive action time (i.e., duration of evasive action) on rear-end collision risk. Method: Fifty drivers’ car-following behavior on a freeway was observed using a driving simulator. An SSM called “Deceleration Rate to Avoid Crash (DRAC)” and the evasive action time were determined for each driver using the data from the driving simulator. Each driver tested two traffic scenarios – Cars and Trucks scenarios where conflicting vehicles were cars and trucks, respectively. The factors related to DRAC were identified and their effects on DRAC were analyzed using the Generalized Linear Models and random effects models. Results: DRAC decreased with the evasive action time and DRAC was closely related to drivers’ gender and driving experience at the road sections where evasive action to avoid collision was required. DRAC was also significantly different between Cars and Trucks scenarios. The effect of the evasive action time on DRAC varied among different drivers, particularly in the Trucks scenario. Conclusions: Longer evasive action time can significantly reduce crash risk. Driver characteristics are more closely related to effective evasive action in complex driving conditions. Practical Applications: Based on the findings of this study, driver warning information can be developed to alert drivers to take specific evasive action that reduces collision risk in a critical traffic event. The information is likely to reduce the variability of the driver’s evasive action and the speed variations among different drivers.     

The detection of drowsiness using a driver monitoring system

Abstract

Objective: Drowsiness is a major cause of driver impairment leading to crashes and fatalities. Research has established the ability to detect drowsiness with various kinds of sensors. We studied drowsy driving in a high-fidelity driving simulator and evaluated the ability of an automotive production-ready driver monitoring system (DMS) to detect drowsy driving. Additionally, this feature was compared to and combined with signals from vehicle-based sensors.

Methods: The National Advanced Driving Simulator was used to expose drivers to long, monotonous drives. Twenty participants drove for about 4?h in the simulator between 10 p.m. and 2 a.m. They were allowed to use cruise control and traffic was sparse and semirandom, with both slower- and faster-moving vehicles. Observational ratings of drowsiness (ORDs) were used as the ground truth for drowsiness, and several dependent measures were calculated from vehicle and DMS signals. Drowsiness classification models were created that used only vehicle signals, only driver monitoring signals, and a combination of the 2 sources.

Results: The model that used DMS signals performed better than the one that used only vehicle signals; however, the combination of the two performed the best. The models were effective at discriminating low levels of drowsiness from moderate to severe drowsiness; however, they were not effective at telling the difference between moderate and severe levels. A binary model that lumped drowsiness into 2 classes had an area under the receiver operating characteristic (ROC) curve of 0.897.

Conclusions: Blinks and saccades have been shown to be predictive of microsleeps; however, it may be that detection of microsleeps and lane departures occurs too late. Therefore, it is encouraging that the model was able to distinguish mild from moderate drowsy driving. The use of automation may make vehicle-based signals useless for characterizing driver states, providing further motivation for a DMS. Future improvements in impairment detection systems may be expected through a combination of improved hardware, physiological measures from unobtrusive sensors and wearables, and the intelligent integration of environmental variables like time of day and time on task.     

The potential of vehicle and road infrastructure interventions in fatal bicyclist accidents on Swedish roads—What can in-depth studies tell us?

Abstract

Objective: The objective of this article is to describe the characteristics of fatal crashes with bicyclists on Swedish roads in rural and urban areas and to investigate the potential of bicycle helmets and different vehicle and road infrastructure interventions to prevent them. The study has a comprehensive approach to provide road authorities and vehicle manufacturers with recommendations for future priorities.

Methods: The Swedish Transport Administration’s (STA) in-depth database of fatal crashes was used for case-by-case analysis of fatal cycling accidents (2006–2016) on rural (n?=?82) and urban (n?=?102) roads. The database consists of information from the police, medical journals, autopsy reports, accident analyses performed by STA, and witness statements. The potential of helmet use and various vehicle and road infrastructure safety interventions was determined retrospectively for each case by analyzing the chain of events leading to the fatality. The potential of vehicle safety countermeasures was analyzed based on prognoses on their implementation rates in the Swedish vehicle fleet.

Results: The most common accident scenario on rural roads was that the bicyclist was struck while cycling along the side of the road. On urban roads, the majority of accidents occurred in intersections. Most accidents involved a passenger car, but heavy trucks were also common, especially in urban areas. Most accidents occurred in daylight conditions (73%). Almost half (46%) of nonhelmeted bicyclists would have survived with a helmet. It was assessed that nearly 60% of the fatal accidents could be addressed by advanced vehicle safety technologies, especially autonomous emergency braking with the ability to detect bicyclists. With regard to interventions in the road infrastructure, separated paths for bicyclists and bicycle crossings with speed calming measures were found to have the greatest safety potential. Results indicated that 91% of fatally injured bicyclists could potentially be saved with known techniques. However, it will take a long time for such technologies to be widespread.

Conclusions: The majority of fatally injured bicyclists studied could potentially be saved with known techniques. A speedy implementation of important vehicle safety systems is recommended. A fast introduction of effective interventions in the road infrastructure is also necessary, preferably with a plan for prioritization.     

High beam headlamp use rates: Effects of rurality,proximity of other traffic,and roadway curvature

Objective: The few observational studies of the prevalence of high beam use indicate the rate of high beam use is about 25% when vehicles are isolated from other vehicles on unlit roads. Recent studies were limited to 2-lane rural roads and used measurement methods that likely overestimated use. The current study examined factors associated with the rate of high beam use of isolated vehicles on a variety of roadways in the Ann Arbor, Michigan area.

Methods: Twenty observation sites were categorized as urban, rural, or on a rural/urban boundary and selected to estimate the effects of street lighting, road curvature, and direction of travel relative to the city on high beam use. Sites were selected in pairs so that a majority of traffic passing one site also passed through the other. Measurement of high beams relied on video data recorded for 2 nights at each site, and the video data also were used to derive a precise measure of the proximity of other traffic. Nearly 3,200 isolated vehicles (10 s or longer from other vehicles) were observed, representing 1,500-plus vehicle pairs.

Results: Across the sample, 18% of the vehicles used high beams. Seventy-three percent of the 1,500-plus vehicle pairs used low beams at each paired site, whereas 9% used high beams at both sites. Vehicles at rural sites and sites at the boundaries of Ann Arbor were more likely to use high beams than vehicles at urban sites, but use in rural areas compared with rural/urban boundary areas did not vary significantly. Rates at all sites were much lower than expected, ranging from 0.9 to 52.9%. High beam use generally increased with greater time between subject vehicles and leading vehicles and vehicles in the opposing lane. There were mixed findings associated with street lighting, road curvature, and direction of travel relative to the city.

Conclusion: Maximizing visibility available to drivers from headlights includes addressing the substantial underuse of high beam headlamps. Advanced technologies such as high beam assist, which switches automatically between high and low beam headlamps depending on the presence of other traffic, can help to address this problem.     

Are older drivers’ driving patterns during an on-road driving task representative of their real-world driving patterns?

Abstract

Objective: The current study investigated whether older drivers’ driving patterns during a customized on-road driving task were representative of their real-world driving patterns.

Methods: Two hundred and eight participants (male: 68.80%; mean age?=?81.52 years, SD?=?3.37 years, range?=?76.00–96.00 years) completed a customized on-road driving task that commenced from their home and was conducted in their own vehicle. Participants’ real-world driving patterns for the preceding 4-month period were also collected via an in-car recording device (ICRD) that was installed in each participant’s vehicle.

Results: During the 4-month period prior to completing the on-road driving task, participants’ median real-world driving trip distance was 2.66?km (interquartile range [IQR]?=?1.14–5.79?km) and their median on-road driving task trip distance was 4.41?km (IQR?=?2.83–6.35?km). Most participants’ on-road driving task trip distances were classified as representative of their real-world driving trip distances (95.2%, n?=?198).

Conclusions: These findings suggest that most older drivers were able to devise a driving route that was representative of their real-world driving trip distance. Future research will examine whether additional aspects of the on-road driving task (e.g., average speed, proportion of trips in different speed zones) are representative of participants’ real-world driving patterns.     

Residual road departure crashes after full deployment of LDW and LDP systems

Abstract

Objective: Road departures are one of the most severe crash modes in the United States. To help reduce this risk, vehicles are being introduced in the United States with lane departure warning (LDW) systems, which warn the driver of a departure, and lane departure prevention (LDP) systems, which assist the driver in steering back to the roadway. Previous studies have estimated that LDW/LDP systems may prevent one third of drift-out-of-lane road departure crashes. This study investigates the crashes that were not prevented, to potentially set research priorities for next-generation road departure prevention systems.

Methods: The event data recorder (EDR) data from 128 road departure crashes in the National Automotive Sampling System Crashworthiness Data System (NASS-CDS) from 2011 to 2015 were mapped onto the vehicle trajectory and simulated with LDW/LDP to assess the potential for crash avoidance. The model predicted that 63–83% of single-vehicle road departure crashes may not be prevented by an LDW system and 49% may not be prevented by an LDP system.

Results and Conclusions: For LDP systems, which were assumed to have zero latency, no crashes were avoided if the time-to-collision (TTC) from lane crossing to impact was less than 0.55?s. Obstacles such as guardrails and traffic barriers, which tend to be very close to the road, were more common among the remaining crashes. The study shows that LDW/LDP systems are limited by two factors, driver reaction time and TTC to the roadside object. Thus, earlier driver response and longer TTC may help in these situations.     

Reducing the risk of rear-end collisions with infrastructure-to-vehicle (I2V) integration of variable speed limit control and adaptive cruise control system

Objective: Adaptive cruise control (ACC) has been investigated recently to explore ways to increase traffic capacity, stabilize traffic flow, and improve traffic safety. However, researchers seldom have studied the integration of ACC and roadside control methods such as the variable speed limit (VSL) to improve safety. The primary objective of this study was to develop an infrastructure-to-vehicle (I2V) integrated system that incorporated both ACC and VSL to reduce rear-end collision risks on freeways.

Methods: The intelligent driver model was firstly modified to simulate ACC behavior and then the VSL strategy used in this article was introduced. Next, the I2V system was proposed to integrate the 2 advanced techniques, ACC and VSL. Four scenarios of no control, VSL only, ACC only, and the I2V system were tested in simulation experiments. Time exposed time to collision (TET) and time integrated time to collision (TIT), 2 surrogate safety measures derived from time to collision (TTC), were used to evaluate safety issues associated with rear-end collisions. The total travel times of each scenario were also compared.

Results: The simulation results indicated that both the VSL-only and ACC-only methods had a positive impact on reducing the TET and TIT values (reduced by 53.0 and 58.6% and 59.0 and 65.3%, respectively). The I2V system combined the advantages of both ACC and VSL to achieve the most safety benefits (reduced by 71.5 and 77.3%, respectively). Sensitivity analysis of the TTC threshold also showed that the I2V system obtained the largest safety benefits with all of the TTC threshold values. The impact of different market penetration rates of ACC vehicles in I2V system indicated that safety benefits increase with an increase in ACC proportions.

Conclusions: Compared to VSL-only and ACC-only scenarios, this integrated I2V system is more effective in reducing rear-end collision risks. The findings of this study provide useful information for traffic agencies to implement novel techniques to improve safety on freeways.     

The Real-World Safety Potential of Connected Vehicle Technology

Objective: This article estimates the safety potential of a current commercially available connected vehicle technology in real-world crashes.

Method: Data from the Centre for Automotive Safety Research's at-scene in-depth crash investigations in South Australia were used to simulate the circumstances of real-world crashes. A total of 89 crashes were selected for inclusion in the study. The crashes were selected as representative of the most prevalent crash types for injury or fatal crashes and had potential to be mitigated by connected vehicle technology. The trajectory, speeds, braking, and impact configuration of the selected in-depth cases were replicated in a software package and converted to a file format allowing “replay” of the scenario in real time as input to 2 Cohda Wireless MK2 onboard units. The Cohda Wireless onboard units are a mature connected vehicle technology that has been used in both the German simTD field trial and the U.S. Department of Transport's Safety Pilot project and have been tuned for low false alarm rates when used in the real world. The crash replay was achieved by replacing each of the onboard unit Global Positioning System (GPS) inputs with the simulated data of each of the involved vehicles. The time at which the Cohda Wireless threat detection software issued an elevated warning was used to calculate a new impact speed using 3 different reaction scenarios and 2 levels of braking.

Results: It was found that between 37 and 86% of the simulated crashes could be avoided, with highest percentage due a fully autonomous system braking at 0.7 g. The same system also reduced the impact speed relative to the actual crash in all cases. Even when a human reaction time of 1.2 s and moderate braking of 0.4 g was assumed, the impact speed was reduced in 78% of the crashes. Crash types that proved difficult for the threat detection engine were head-on crashes where the approach angle was low and right turn–opposite crashes.

Conclusions: These results indicate that connected vehicle technology can be greatly beneficial in real-world crash scenarios and that this benefit would be maximized by having the vehicle intervene autonomously with heavy braking. The crash types that proved difficult for the connected vehicle technology could be better addressed if controller area network (CAN) information is available, such as steering wheel angle, so that driver intent can be inferred sooner. More accurate positioning in the real world (e.g., combining satellite positioning and accelerometer data) would allow the technology to be more effective for near-collinear head-on and rear-end crashes, because the low approach angles that are common in such crashes are currently ignored in order to minimize false alarms due to positioning uncertainty.