Crash prediction with behavioral and physiological features for advanced vehicle collision avoidance system

Real-time crash prediction is the key component of the Vehicle Collision Avoidance System (VCAS) and other driver assistance systems. The further improvements of predictability requires the systemic estimation of crash risks in the driver-vehicle-environment loop. Therefore, this study designed and validated a prediction method based on the supervised learning model with added behavioral and physiological features. The data samples were extracted from 130 drivers’ simulator driving, and included various features generated from synchronized recording of vehicle dynamics, distance metrics, driving behaviors, fixations and physiological measures. In order to identify the optimal configuration of proposed method, the Discriminant Analysis (DA) with different features and models (i.e. linear or quadratic) was tested to classify the crash samples and non-crash samples. The results demonstrated the significant improvements of accuracy and specificity with added visual and physiological features. The different models also showed significant effects on the characteristics of sensitivity and specificity. These results supported the effectiveness of crash prediction by quantifying drivers’ risky states as inputs. More importantly, such an approach also provides opportunities to integrate the driver state monitoring into other vehicle-mounted systems at the software level.     

Do cooperative systems make drivers’ car-following behavior safer?

The main goal of in-vehicle technologies and co-operative services is to reduce congestion and increase traffic safety. This is achieved by alerting drivers on risky traffic conditions ahead of them and by exchanging traffic and safety related information for the particular road segment with nearby vehicles. Road capacity, level of service, safety, and air pollution are impacted to a large extent by car-following behavior of drivers. Car-following behavior is an essential component of micro-simulation models. This paper investigates the impact of an infrastructure-to-vehicle (I2V) co-operative system on drivers’ car-following behavior. Test drivers in this experiment drove an instrumented vehicle with and without the system. Collected trajectory data of the subject vehicle and the vehicle in front, as well as socio-demographic characteristics of the test drivers were used to estimate car-following models capturing their driving behavior with and without the I2V system. The results show that the co-operative system harmonized the behavior of drivers and reduced the range of acceleration and deceleration differences among them. The observed impact of the system was largest on the older group of drivers.     

Analysis of volatility in driving regimes extracted from basic safety messages transmitted between connected vehicles

Driving volatility captures the extent of speed variations when a vehicle is being driven. Extreme longitudinal variations signify hard acceleration or braking. Warnings and alerts given to drivers can reduce such volatility potentially improving safety, energy use, and emissions. This study develops a fundamental understanding of instantaneous driving decisions, needed for hazard anticipation and notification systems, and distinguishes normal from anomalous driving. In this study, driving task is divided into distinct yet unobserved regimes. The research issue is to characterize and quantify these regimes in typical driving cycles and the associated volatility of each regime, explore when the regimes change and the key correlates associated with each regime. Using Basic Safety Message (BSM) data from the Safety Pilot Model Deployment in Ann Arbor, Michigan, two- and three-regime Dynamic Markov switching models are estimated for several trips undertaken on various roadway types. While thousands of instrumented vehicles with vehicle to vehicle (V2V) and vehicle to infrastructure (V2I) communication systems are being tested, nearly 1.4 million records of BSMs, from 184 trips undertaken by 71 instrumented vehicles are analyzed in this study. Then even more detailed analysis of 43 randomly chosen trips (N = 714,340 BSM records) that were undertaken on various roadway types is conducted. The results indicate that acceleration and deceleration are two distinct regimes, and as compared to acceleration, drivers decelerate at higher rates, and braking is significantly more volatile than acceleration. Different correlations of the two regimes with instantaneous driving contexts are explored. With a more generic three-regime model specification, the results reveal high-rate acceleration, high-rate deceleration, and cruise/constant as the three distinct regimes that characterize a typical driving cycle. Moreover, given in a high-rate regime, drivers’ on-average tend to decelerate at a higher rate than their rate of acceleration. Importantly, compared to cruise/constant regime, drivers’ instantaneous driving decisions are more volatile both in “high-rate” acceleration as well as “high-rate” deceleration regime. The study contributes to analyzing volatility in short-term driving decisions, and how changes in driving regimes can be mapped to a combination of local traffic states surrounding the vehicle.     

Developing freeway lane‐changing support systems using artificial neural networks

Lane‐changing involves many concerns about safety and efficiency which makes it one of the most difficult tasks of driving. It is indeed quite personal since drivers operate vehicles according to their integrated perception of comprehensive circumstances rather than individual rules. A lane‐changing decision support model is developed in this study using artificial neural networks (ANN). The advantages of the ANN approach lie in the learning capability. Due to its nature, an ANN model can consolidate various kinds of information surrounding the vehicle for the drivers and generate reliable results to help control vehicles. It then becomes a useful mechanism to assist drivers in judging current situations and making the right decisions. Several preliminary validations and comparisons are conducted with the field survey data. It is confirmed that the ANN model mimics traffic characteristics more accurately than conventional methods. This product would expedite the implementation of relevant applications in the intelligent transportation systems context. In particular, the ANN model can be adapted to individual driver characteristics. This reveals practical feasibility and significant market potential for customized in‐vehicle equipment.     

Using on-board logging devices to study the longer-term impact of an eco-driving course

In this paper the long-term impact of an eco-driving training course is evaluated by monitoring driving behavior and fuel consumption for several months before and after the course. Cars were equipped with an on-board logging device that records the position and speed of the vehicle using GPS tracking as well as real time as electronic engine data extracted from the controller area network. The data includes mileage, number of revolutions per minute, position of the accelerator pedal, and instantaneous fuel consumption. It was gathered over a period of 10 months for 10 drivers during real-life conditions thus enabling an individual drive style analysis. The average fuel consumption four months after the course fell by 5.8%. Most drivers showed an immediate improvement in fuel consumption that was stable over time, but some tended to fall back into their original driving habits.     

Modeling lane-changing behavior in a connected environment: A game theory approach

Vehicle-to-Vehicle communications provide the opportunity to create an internet of cars through the recent advances in communication technologies, processing power, and sensing technologies. A connected vehicle receives real-time information from surrounding vehicles; such information can improve drivers’ awareness about their surrounding traffic condition and lead to safer and more efficient driving maneuvers. Lane-changing behavior, as one of the most challenging driving maneuvers to understand and to predict, and a major source of congestion and collisions, can benefit from this additional information. This paper presents a lane-changing model based on a game-theoretical approach that endogenously accounts for the flow of information in a connected vehicular environment. A calibration approach based on the method of simulated moments is presented and a simplified version of the proposed framework is calibrated against NGSIM data. The prediction capability of the simplified model is validated. It is concluded the presented framework is capable of predicting lane-changing behavior with limitations that still need to be addressed. Finally, a simulation framework based on the fictitious play is proposed. The simulation results revealed that the presented lane-changing model provides a greater level of realism than a basic gap-acceptance model.     

A rule-based neural network approach to model driver naturalistic behavior in traffic

This paper proposes a rule-based neural network model to simulate driver behavior in terms of longitudinal and lateral actions in two driving situations, namely car-following situation and safety critical events. A fuzzy rule based neural network is constructed to obtain driver individual driving rules from their vehicle trajectory data. A machine learning method reinforcement learning is used to train the neural network such that the neural network can mimic driving behavior of individual drivers. Vehicle actions by neural network are compared to actions from naturalistic data. Furthermore, this paper applies the proposed method to analyze the heterogeneities of driving behavior from different drivers’ data.Driving data in the two driving situations are extracted from Naturalistic Truck Driving Study and Naturalistic Car Driving Study databases provided by the Virginia Tech Transportation Institute according to pre-defined criteria. Driving actions were recorded in instrumented vehicles that have been equipped with specialized sensing, processing, and recording equipment.     

Modeling time-of-day car use behavior: A Bayesian network approach

In this research, a Bayesian network (BN) approach is proposed to model the car use behavior of drivers by time of day and to analyze its relationship with driver and car characteristics. The proposed BN model can be categorized as a tree-augmented naive (TAN) Bayesian network. A latent class variable is included in this model to describe the unobserved heterogeneity of drivers. Both the structure and the parameters are learned from the dataset, which is extracted from GPS data collected in Toyota City, Japan. Based on inferences and evidence sensitivity analysis using the estimated TAN model, the effects of each single observed characteristic on car use measures are tested and found to be significant. The features of each category of the latent class are also analyzed. By testing the effect of each car use measure on every other measure, it is found that the correlations between car use measures are significant and should be considered in modeling car use behavior.     

The joint analysis of injury severity of drivers in two-vehicle crashes accommodating seat belt use endogeneity

The current study contributes to the existing injury severity modeling literature by developing a multivariate probit model of injury severity and seat belt use decisions of both drivers involved in two-vehicle crashes. The modeling approach enables the joint modeling of the injury severity of multiple individuals involved in a crash, while also recognizing the endogeneity of seat belt use in predicting injury severity levels as well as accommodating unobserved heterogeneity in the effects of variables. The proposed model is applied to analyze the injury severity of drivers involved in two-vehicle road crashes in Denmark.The empirical analysis provides strong support for the notion that people offset the restraint benefits of seat belt use by driving more aggressively. Also, men and those individuals driving heavy vehicles have a lower injury risk than women and those driving lighter vehicles, respectively. At the same time, men and individuals driving heavy vehicles pose more of a danger to other drivers on the roadway when involved in a crash. Other important determinants of injury severity include speed limit on roadways where crash occurs, the presence (or absence) of center dividers (median barriers), and whether the crash involves a head-on collision. These and other results are discussed, along with implications for countermeasures to reduce injury severities in crashes. The analysis also underscores the importance of considering injury severity at a crash level, while accommodating seat belt endogeneity effects and unobserved heterogeneity effects.     

Modelling dual carriageway lane changing using neural networks

Neural networks offer a potential alternative method of modelling driver behaviour within road traffic systems. This paper explores the application of neural networks to modelling the lane-changing decisions of drivers on dual carriageways. Two approaches are considered. The first, preliminary approach uses a prediction type of neural network with a single hidden layer and the back propagation learning algorithm to model the behaviour of an individual driver. A series of consecutive time-scan traffic patterns, which describe the driver's environment and changes over time as the selected vehicle travels along a link, are input to the neural network, which then predicts the new lane and position of the vehicle. Training data are collected from a human subject using an interactive driving simulation. The trained neural network successfully exhibited the rudiments of driving behaviour in terms of lane and speed changes. A major disadvantage of this approach was the difficulty in recording real-life data, which are required to train the neural network, for individual drivers. The second approach concentrates specifically on lane changing and makes use of a learning vector quantization classification type of neural network. Input to the neural network still consists primarily of time-scan traffic patterns, but the format is changed to facilitate the possibility of data acquisition using image processing. The neural network output classifies the input data by determining the new lane for the vehicle concerned. Performance in both testing and training was very good for data generated by the rule-based driver-decision model of a microscopic simulation. Performance in testing was less satisfactory for data taken directly from a road and highlighted the need for extensive data sets for successful training.     

Effects of crash warning systems on rear-end crash avoidance behavior under fog conditions

Reduced visibility conditions increase both the probability of rear-end crash occurrences and their severity. Crash warning systems that employ data from connected vehicles have potential to improve vehicle safety by assisting drivers to be aware of the imminent situations ahead in advance and then taking timely crash avoidance action(s). This study provides a driving simulator study to evaluate the effectiveness of the Head-up Display warning system and the audio warning system on drivers’ crash avoidance performance when the leading vehicle makes an emergency stop under fog conditions. Drivers’ throttle release time, brake transition time, perception response time, brake reaction time, minimum modified time-to-collision, and maximum brake pedal pressure are assessed for the analysis. According to the results, the crash warning system can help decrease drivers’ reaction time and reduce the probability of rear-end crashes. In addition, the effects of fog level and drivers’ characteristics including gender and age are also investigated in this study. The findings of this study are helpful to car manufacturers in designing rear-end crash warning systems that enhance the effectiveness of the system’s application under fog conditions.     

Modeling of individual vehicle safety and fuel consumption under comprehensive external conditions

This research intends to explore external factors affecting driving safety and fuel consumption, and build a risk and fuel consumption prediction model for individual drivers based on natural driving data. Based on 120 taxi drivers’ natural driving data during 4 months, driving behavior data under various conditions of the roadway, traffic, weather, and time of day are extracted. The driver's fuel consumption is directly collected by the on-board diagnostics (OBD) unit, and safety index is calculated based on Data Threshold Violations (DTV) and Phase Plane Analysis with Limits (PPAL) considering speed, longitudinal and lateral acceleration. By using a linear mixed model explaining the fixed effect of the external conditions and the random effect of the driver, the influences of various external factors on fuel consumption and safety are analyzed and discussed. The prediction model lays a foundation for drivers' fuel consumption and risk prediction in different external conditions, which could help improve individual driving behavior for the benefit of both fuel consumption and safety.     

Analysis of three independent real-world driving studies: A data driven and expert analysis approach to determining parameters affecting fuel economy

It is well established that individual variations in driving style have a significant impact on vehicle energy efficiency. The literature shows certain parameters have been linked to good fuel economy, specifically acceleration, throttle use, number of stop/starts and gear change behaviours. The primary aim of this study was to examine what driving parameters are specifically related to good fuel economy using a non-homogeneous extended data set of vehicles and drivers over real-world driving scenarios spanning two countries. The analysis presented in this paper shows how three completely independent studies looking at the same factor (i.e., the influence of driver behaviour on fuel efficiency) can be evaluated, and, despite their notable differences in location, environment, route, vehicle and drivers, can be compared on broadly similar terms. The data from the three studies were analysed in two ways; firstly, using expert analysis and the second a purely data driven approach. The various models and experts concurred that a combination of at least one factor from the each of the categories of vehicle speed, engine speed, acceleration and throttle position were required to accurately predict the impact on fuel economy. The identification of standard deviation of speed as the primary contributing factor to fuel economy, as identified by both the expert and data driven analysis, is also an important finding. Finally, this study has illustrated how various seemingly independent studies can be brought together, analysed as a whole and meaningful conclusions extracted from the combined data set.     

A driving cycle detection approach using map service API

Following advancements in smartphone and portable global positioning system (GPS) data collection, wearable GPS data have realized extensive use in transportation surveys and studies. The task of detecting driving cycles (driving or car-mode trajectory segments) from wearable GPS data has been the subject of much research. Specifically, distinguishing driving cycles from other motorized trips (such as taking a bus) is the main research problem in this paper. Many mode detection methods only focus on raw GPS speed data while some studies apply additional information, such as geographic information system (GIS) data, to obtain better detection performance. Procuring and maintaining dedicated road GIS data are costly and not trivial, whereas the technical maturity and broad use of map service application program interface (API) queries offers opportunities for mode detection tasks. The proposed driving cycle detection method takes advantage of map service APIs to obtain high-quality car-mode API route information and uses a trajectory segmentation algorithm to find the best-matched API route. The car-mode API route data combined with the actual route information, including the actual mode information, are used to train a logistic regression machine learning model, which estimates car modes and non-car modes with probability rates. The experimental results show promise for the proposed method’s ability to detect vehicle mode accurately.     

Naturalistic driving data collection to investigate into the effects of road geometrics on track behaviour

Road designers assume that drivers will follow the road alignment with trajectories centred in the lane, and move at the design speed parallel to the road centreline (i.e., the horizontal alignment). Therefore, they assume that if the horizontal alignment indicates the “designed trajectory”, the driving path indicates the “operating trajectory”. However, at present, they do not have the necessary tools to measure the relationship between the designed alignment and possible vehicle trajectories.The paper has two objectives: (a) to develop an understanding of the root causes of differences between road alignment and vehicle trajectories; and (b) to define and calibrate a model that estimates the local curvature of trajectories on the basis of the designed horizontal alignment.The two objectives were pursued by carrying out a naturalistic survey using vehicles equipped with high precision GPS in real-time kinematics (RTK) mode driven by test drivers on road sections of known geometric characteristics. The results provide an insight into the effects of road geometrics on driver behaviour, thus anticipating possible driving errors or unexpected/undesired behaviours, information which can then be used to correct possible inconsistencies when making decisions at the design stage.     

Modeling traffic's flow-density relation: Accommodation of multiple flow regimes and traveler types

This research investigates freeway-flow impacts of different traveler types by specifying and applying a latent-segmentation model of congested and uncongested driving behaviors. Drivers in uncongested conditions are assumed to drive at self-chosen speeds, while drivers in congested conditions are assumed to take speed as given and choose a spacing (between their vehicle and the previous vehicle). Several classes of driver-vehicle combinations are distinguished in a data set based on double-loop-detector pulses and a household travel survey. These classifications are made on the basis of vehicle type and gender, leading to class estimates of speeds and spacings. The segmentation model is specified as a logit function of density, weather, and vehicle type, leading to estimates of congested-condition probabilities. Unobserved heterogeneity is incorporated in all models via common error assumptions.Results indicate that segmentation models are promising tools for traffic data analysis and that information on travelers, their vehicles, and weather conditions explains significant variation in flow data. By clarifying a greater understanding of traffic conditions and traveler behavior explains much scatter in the fundamental relation between flow, speed, and density, can assist regions in their traffic-management efforts and engineers in their design of roadway facilities. Ultimately, such improvements to travel networks should enhance quality of life.     

Implications of driving style and road grade for accurate vehicle activity data and emissions estimates

Real-world vehicle operating mode data (2.5 million 1 Hz records), collected by instrumenting the vehicles of 82 volunteer drivers with OBD datalogger and GPS while they drove their routine travel routes, were analyzed to quantify vehicle emissions estimate errors due to road grade and driving style in rural, hilly Vermont. Data were collected in winter and summer for MY 1996 and newer passenger cars and trucks only. EPA MOVES2010b was used to estimate running exhaust emissions associated with measured vehicle activity. Changes in vehicle specific power (VSP) and MOVES operating mode (OpMode) due to proper accounting for real-world road grade indicated emission rate errors between 10% and 48%, depending on pollutant, chiefly because grade-related changes in VSP could shift activity by as many as six OpModes, depending on road type. The correct MOVES OpMode assignment was made only 33–55% of the time when road grade was not included in the VSP calculation. Driving style of individual drivers was difficult to assess due to unknown traffic operations data, but the largest differences between individual drivers were observed on rural restricted roads, where traffic conditions and control have minimal impact. The results suggest the importance of (1) measuring and incorporating real-world road grade in order to correctly assign MOVES emission rates; and (2) developing a driving style typology to account for differences in the MOVES emissions estimates due to driver variability.     

Identification of freeway crash-prone traffic conditions for traffic flow at different levels of service

The primary objective of this study was to evaluate the risks of crashes associated with the freeway traffic flow operating at various levels of service (LOS) and to identify crash-prone traffic conditions for each LOS. The results showed that the traffic flow operating at LOS E had the highest crash potential, followed by LOS F and D. The traffic flow operating at LOS B and A had the lowest crash potential. For LOS A and B, the vehicle platoon and abrupt change in vehicle speeds were major contributing factors to crash occurrences. For LOS C, crash risks were correlated with lane-change maneuvers, speed variation, and small headways in traffic. For LOS D, crash risks increased with an increase in the temporal change in traffic flow variables and the frequency of lane-change maneuvers. For LOS E, crash risks were mainly affected by high traffic volumes and oscillating traffic conditions. For LOS F, crash risks increased with an increase in the standard deviation of flow rate and the frequency of lane-change maneuvers. The findings suggested that the mechanism of crashes were quite different across various LOS. A Bayesian random-parameters logistic regression model was developed to identify crash-prone traffic conditions for various LOS. The proposed model significantly improved the prediction performance as compared to the conventional logistic regression model.