Safety Pilot Model Deployment Research Articles

Modeling coupled driving behavior during lane change: A multi-agent Transformer reinforcement learning approach

In a lane change (LC) scenario, the lane change vehicle interacts with surrounding vehicles. The interactions not only affect their driving behaviors but also influence the traffic flow. This study aims to model the coupled behavior of the lane changer and the follower in the target lane during LC. Large-scale real-world connected vehicle (CV) data from the Safety Pilot Model Deployment (SPMD) program are used to extract LCs and study vehicle interactions. A multi-agent Transformer-based deep deterministic policy gradient (MA-TDDPG) method is proposed to model the coupled behaviors during LC. The multi-agent framework can handle the multiple agents’ behaviors with interactions, and the Transformer can process the observation-action memory accurately and efficiently. The MA-TDDPG algorithm can learn the sequential decision-making process over continuous action space during LC with the accommodation of the multi-vehicle interaction and the driver’s memory effect. Compared to traditional supervised learning and reinforcement learning methods, it demonstrates superior performance in imitating the longitudinal and lateral actions of the lane changer and the follower. The findings of this study provide insights into the development of microscopic simulations by producing realistic LC behaviors, and assistance/automation LC systems by generating LC motions and responses conforming to human driving habits. The model also creates an interactive simulation environment and lays the foundation for optimizing driving strategies.

Application of CNN-LSTM Model for Vehicle Acceleration Prediction Using Car-following Behavior Data

Accurate vehicle acceleration prediction is useful for developing reliable Advanced Driving Assistance Systems (ADAS) and improving road safety. The existence of driver heterogeneity magnifies the variations in acceleration data, leading to consequential impacts on the precision of vehicle acceleration prediction. However, few studies have fully considered the driver heterogeneity when predicting vehicle acceleration. To model the characteristics of individual drivers, this study first identifies the driving behavior semantics which is defined as the underlying patterns of driving behaviors. The analysis results from the coupled hidden Markov model (CHMM) are used to evaluate the driving behavior differences between different drivers by Wasserstein distance. Then the convolutional neural network (CNN) and long short-term memory (LSTM) network are applied to predict vehicle acceleration. To validate the accuracy of the proposed prediction framework, vehicle acceleration data in car-following conditions is extracted from the safety pilot model deployment (SPMD) dataset. The segmentation results indicate that the CHMM possesses a robust capacity for modeling driving behavior. The prediction results demonstrate that the proposed framework, which incorporates driver clustering before prediction, significantly improves the accuracy of predictions. And the CNN-LSTM outperforms the LSTM in predicting vehicle acceleration during car-following scenarios. The findings from this study can enhance the development of personalized functionalities within ADAS to promote its deployment, thereby improving its acceptance and safety.

Modeling driver’s evasive behavior during safety–critical lane changes: Two-dimensional time-to-collision and deep reinforcement learning

Lane changes are complex driving behaviors and frequently involve safety–critical situations. This study aims to develop a lane-change-related evasive behavior model, which can facilitate the development of safety-aware traffic simulations and predictive collision avoidance systems. Large-scale connected vehicle data from the Safety Pilot Model Deployment (SPMD) program were used for this study. A new surrogate safety measure, two-dimensional time-to-collision (2D-TTC), was proposed to identify the safety–critical situations during lane changes. The validity of 2D-TTC was confirmed by showing a high correlation between the detected conflict risks and the archived crashes. A deep deterministic policy gradient (DDPG) algorithm, which could learn the sequential decision-making process over continuous action spaces, was used to model the evasive behaviors in the identified safety–critical situations. The results showed the superiority of the proposed model in replicating both the longitudinal and lateral evasive behaviors.

Predicting intersection crash frequency using connected vehicle data: A framework for geographical random forest

Accurate crash frequency prediction is critical for proactive safety management. The emerging connected vehicles technology provides us with a wealth of vehicular motion data, which enables a better connection between crash frequency and driving behaviors. However, appropriately dealing with the spatial dependence of crash frequency and multitudinous driving features has been a difficult but critical challenge in the prediction process. To this end, this study aims to investigate a new Artificial Intelligence technique called Geographical Random Forest (GRF) that can address spatial heterogeneity and retain all potential predictors. By harnessing more than 2.2 billion high-resolution connected vehicle Basic Safety Message (BSM) observations from the Safety Pilot Model Deployment in Ann Arbor, MI, 30 indicators of driving volatility are extracted, including speed, longitudinal and lateral acceleration, and yaw rate. The developed GRF was implemented to predict rear-end crash frequency at intersections. The results show that: 1) rear-end crashes are more likely to happen at intersections connecting minor roads compared to major roads; 2) a higher number of hard acceleration and deceleration events beyond two standard deviations in the longitudinal direction is a leading indicator of rear-end crashes; 3) the optimal GRF significantly outperforms Global Random Forest, with a 9% lower test error and a substantially better fit; and 4) geographical visualization of variable importance highlights the presence of spatial non-stationarity. The proposed framework can proactively identify at-risk intersections and alert drivers when leading indicators of driving volatility tend to worsen.

Incorporating driving volatility measures in safety performance functions: Improving safety at signalized intersections

About 40 percent of motor vehicle crashes in the US are related to intersections. To deal with such crashes, Safety Performance Functions (SPFs) are vital elements of the predictive methods used in the Highway Safety Manual. The predictions of crash frequencies and potential reductions due to countermeasures are based on exposure and geometric variables. However, the role of driving behavior factors, e.g., hard accelerations and declarations at intersections, which can lead to crashes, are not explicitly treated in SPFs. One way to capture driving behavior is to harness connected vehicle data and quantify performance at intersections in terms of driving volatility measures, i.e., rapid changes in speed and acceleration. According to recent studies, driving volatility is typically associated with higher risk and safety-critical events and can serve as a surrogate for driving behavior. This study incorporates driving volatility measures in the development of SPFs for four-leg signalized intersections. The Safety Pilot Model Deployment (SPMD) data containing over 125 million Basic Safety Messages generated by over 2,800 connected vehicles are harnessed and linked with the crash, traffic, and geometric data belonging to 102 signalized intersections in Ann Arbor, Michigan. The results show that including driving volatility measures in SPFs can reduce model bias and significantly enhances the models' goodness-of-fit and predictive performance. Technically, the best results were obtained by applying Bayesian hierarchical Negative Binomial Models, which account for spatial correlation between signalized intersections. The results of this study have implications for practitioners and transportation agencies about incorporating driving behavior factors in the development of SPFs for greater accuracy and measures that can potentially reduce volatile driving.

Multivariate analysis of car-following behavior data using a coupled hidden Markov model

Accurate analysis of driving behavior data is important for improving traffic operations and for automakers to design safe and effective Advanced Driver Assistance System (ADAS). Although different approaches have been proposed to analyze driving behavior data and understand the car-following behavior, very few of them have considered the interaction effects among driving behavior variables and constructed a multivariate modeling framework to fully understand and extract the underlying primitive driving patterns. To bridge this gap, this paper introduces the coupled hidden Markov model (CHMM) and evaluates its applicability in obtaining the car-following behavioral semantics and comparing the similarity of different drivers’ driving styles. The Safety Pilot Model Deployment (SPMD) dataset, which contains two years of naturalistic driving data, is used to demonstrate the advantages of the proposed coupled modeling approach. The modeling results suggest that CHMM with 3 hidden states provides the most appropriate driving behavior pattern segmentation results. It generates reasonable durations for each behavioral semantic by accounting for the dependence relationship among variables in the analysis. Moreover, a similarity analysis is conducted considering three aspects of a driving behavior pattern: driving behavior pattern transfer, driving behavior pattern selection, and aggressiveness behavior. The result reveals that these three aspects do not necessarily show consistent trends for the same driving behavior. Overall, the findings from this paper provide a novel framework for better understanding and interpreting the driving behavior and primitive driving patterns, and confirm the importance of considering the dependence among different driving behavior variables.

Sensor Data Based Anomaly Detection in Autonomous Vehicles using Modified Convolutional Neural Network

Automated Vehicles (AVs) reform the automotive industry by enabling real-time and efficient data exchange between the vehicles. While connectivity and automation of the vehicles deliver a slew of benefits, they may also introduce new safety, security, and privacy risks. Further, AVs rely entirely on the sensor data and the data from other vehicles too. On the other hand, the sensor data is susceptible to anomalies caused by cyber-attacks, errors, and faults, resulting in accidents and fatalities. Hence, it is essential to create techniques for detecting anomalies and identifying their sources before the wide adoption of AVs. This paper proposes an anomaly detection model using a Modified-Convolutional Neural Network (M-CNN) with Safety Pilot Model Deployment (SPMD) dataset. The M-CNN model comprises specifically trained layers involving the ReLU activation function for feature extraction and detection of AV anomalies. Furthermore, the Adam is used as the optimization algorithm to train the model. The detection accuracy of the proposed model is compared with Isolation Forest (IF) and Support Vector Machine (SVM). The experimental result reveals that the proposed model outperforms the other models with an accuracy of 99.40% in AV anomaly detection.

Driving behavior at midblock crosswalks with Rectangular Rapid Flashing Beacons: Hidden Markov model approach using naturalistic data

Pedestrian fatalities have increased in the U.S. with the largest increase being observed on urban arterials and away from intersections. Rectangular Rapid Flashing Beacon (RRFB) has been widely implemented as a safety countermeasure to improve driver awareness and visibility of pedestrians, especially for midblock crosswalks. Studies show that drivers are more likely to yield to pedestrians at crosswalks with an RRFB. These studies are often based on a binary outcome of whether or not drivers yield to pedestrians. Nevertheless, it is also important to consider the drivers’ deceleration behavior as a dynamic process at these crosswalks and the impact of pedestrians being present or not. Understanding this dynamic behavior and the related circumstances can provide information on the design of alerting systems that help drivers make more appropriate decisions at these crosswalks to avoid a vehicle–pedestrian crash. This study examined this research topic using Hidden Markov Models (HMMs) and data from a naturalistic study. More specifically, four HMMs were applied to the naturalistic brake and jerk data from the Safety Pilot Model Deployment (SPMD) program given drivers’ intention to slow down, the RRFB activation status, and the presence of pedestrians. The time-based data sequence was converted to distance-based through a moving window to enhance result comparison and interpretation. Grid-search was used to select the best moving window parameters and the optimal number of hidden states. This study confirmed the high compliance at an activated RRFB when pedestrians were present. Even without pedestrians, one in five traversals showed drivers slowing down to less than 8.94 m/s (20 mph) within 35 m of the crosswalk. Model results further indicate that drivers started braking as far back as 180 m before the crosswalk and stopped braking from 70 m before the crosswalk at an activated RRFB without pedestrians. When there were pedestrians, drivers would start braking 20 to 30 m later but would brake more firmly and for longer. Finally, drivers were not likely to brake or decelerate when RRFB was off and no pedestrians were present.

Learning-Based Stochastic Driving Model for Autonomous Vehicle Testing

In the simulation-based testing and evaluation of autonomous vehicles (AVs), how background vehicles (BVs) drive directly influences the AV’s driving behavior and further affects the test results. Most existing simulation platforms use either predetermined trajectories or deterministic driving models to model BV behaviors. However, predetermined BV trajectories cannot react to AV maneuvers, and deterministic models are different from real human drivers because of the lack of stochastic components and errors. Both methods lead to unrealistic traffic scenarios. This paper presents a learning-based stochastic driving model that meets the unique needs of AV testing (i.e., interactive and human-like stochasticity). The model is built based on the long short-term memory architecture. By incorporating the concept of quantile regression into the loss function of the model, the stochastic behaviors are reproduced without prior assumption of human drivers. The model is trained with the large-scale naturalistic driving data (NDD) from the Safety Pilot Model Deployment project and compared with a stochastic intelligent driving model (IDM). Analysis of individual trajectories shows that the proposed model can reproduce more similar trajectories of human drivers than IDM. To validate the ability of the proposed model in generating a naturalistic driving environment, traffic simulation experiments are implemented. The results show that traffic flow parameters such as speed, range, and time headway distribution match closely with the NDD, which is of significant importance for AV testing and evaluation.

Online distraction detection for naturalistic driving dataset using kinematic motion models and a multiple model algorithm

Detecting distracted driving is important for developing Advanced Driver Assistance Systems and improving road safety. Most of the existing research analyzes drivers directly via video analysis techniques or by measuring cognitive load, however these approaches often require additional sensors to be installed in vehicles or equipped to drivers. Given that most distractions may have a direct influence on drivers’ control of vehicles, this paper proposes a new method to utilize available vehicle kinematic data for detecting distracted driving. The proposed method predicts vehicle kinematics by fusing multiple state–space models that capture different driving motion patterns under normal driving. An online monitoring scheme is developed by using Exponentially Weighted Moving Average (EWMA) and Cumulative Sum (CUSUM) charts, which detects abnormal mean shifts of lateral speeds and prediction errors of lane positions to provide warnings of distracted driving. A case study is presented based on two naturalistic driving datasets — the Integrated Vehicle-Based Safety Systems (IVBSS) and Safety Pilot Model Deployment (SPMD) datasets.

Driving Style Recognition under Connected Circumstance Using a Supervised Hierarchical Bayesian Model

In recent years, the automated driving system has been known to be one of the most popular research topics of artificial intelligence (AI) and intelligent transportation system (ITS). The journey experience on automated vehicles and the intelligent automated driving system could be improved by individualization driving understanding. Although previous studies have proposed methods for driving styles understanding, the individualization driving classification has not been addressed thoroughly. Therefore, in this study, a supervised method is proposed to understand driving behavioral structure and the latent driving styles by incorporating the prior knowledge. Firstly, a novel method is established for driving behavioral encoding and raw driving data mining. Then, the Labeled Latent Dirichlet Allocation (LLDA) is proposed to understand the latent driving styles from individual driving with driving behaviors. Finally, the Safety Pilot Model Deployment (SPMD) data are used to validate the performance of the proposed model. Experimental results show that the proposed model uncovers latent driving styles effectively and shows good agreement to real situations, which provides theoretical guidance on driving behavior recognition for better individual experience on automated driving vehicles.

Hidden Markov Model of Lane-Changing-Based Car-Following Behavior on Freeways using Naturalistic Driving Data

This paper develops a methodology for simultaneously modeling lane-changing and car-following behavior of automated vehicles on freeways. Naturalistic driving data from the Safety Pilot Model Deployment (SPMD) program are used. First, a framework to process the SPMD data is proposed using various data analytics techniques including data fusion, data mining, and machine learning. Second, pairs of automated host vehicle and their corresponding front vehicle are identified along with their lane-change and car-following relationship data. Using these data, a lane-changing-based car-following (LCCF) model, which explicitly considers lane-change and car-following behavior simultaneously, is developed. The LCCF model is based on Gaussian-mixture-based hidden Markov model theory and is disaggregated into two processes: LCCF association and LCCF dissociation. These categories are based on the result of the lane change. The overall goal is to predict a driver’s lane-change intention using the LCCF model. Results show that the model can predict the lane-change event in the order of 0.6 to 1.3 s before the moment of the vehicle body across the lane boundary. In addition, the execution times of lane-change maneuvers average between 0.55 and 0.86 s. The LCCF model allows the intention time and execution time of driver’s lane-change behavior to be forecast, which will help to develop better advanced driver assistance systems for vehicle controls with respect to lane-change and car-following warning functions.

The role of drivers’ social interactions in their driving behavior: Empirical evidence and implications for car-following and traffic flow

Models for describing the microscopic driving behavior rarely consider the “social effects” on drivers’ driving decisions. However, social effect can be generated due to interactions with surrounding vehicles and affect drivers’ driving behavior, e.g., the interactions result in imitating the behavior of peer drivers. Therefore, social environment and peer influence can impact the drivers’ instantaneous behavior and shift the individuals’ driving state. This study aims to explore empirical evidence for existence of a social effect, i.e., when a fast-moving vehicle passes a subject vehicle, does the driver mimic the behavior of passing vehicle? High-resolution Basic Safety Message data set (N = 151,380,578) from the Safety Pilot Model Deployment program in Ann Arbor, Michigan, is used to explore the issue. The data relates to positions, speeds, and accelerations of 63 host vehicles traveling in connected vehicles with detailed information on surrounding environment at a frequency of 10 Hz. Rigorous random parameter logit models are estimated to capture the heterogeneity among the observations and to explore if the correlates of social effect can vary both positively and negatively. Results show that subject drivers do mimic the behavior of passing vehicles –in 16 percent of passing events (N = 18,099 total passings occurred in freeways), subject vehicle drivers are observed to follow the passing vehicles accelerating. We found that only 1.2 percent of drivers normally sped up (10 km/hr in 10 s) during their trips, when they were not passed by other vehicles. However, if passed by a high speed vehicle the percentage of drivers who sped up is 16.0 percent. The speed change of at least 10 km/hr within 10 s duration is considered as accelerating threshold. Furthermore, the acceleration of subject vehicle is more likely if the speed of subject driver is higher and more surrounding vehicles are present. Interestingly, if the difference with passing vehicle speed is high, the likelihood of subject driver’s acceleration is lower, consistent with expectation that if such differences are too high, the subject driver may be minimally affected. The study provides new evidence that drivers’ social interactions can change traffic flow and implications of the study results are discussed.

Online predictive connected and automated eco-driving on signalized arterials considering traffic control devices and road geometry constraints under uncertain traffic conditions

For energy-efficient Connected and Automated Vehicle (CAV) Eco-driving control on signalized arterials under uncertain traffic conditions, this paper explicitly considers traffic control devices (e.g., road markings, traffic signs, and traffic signals) and road geometry (e.g., road shapes, road boundaries, and road grades) constraints in a data-driven optimization-based Model Predictive Control (MPC) modeling framework. This modeling framework uses real-time vehicle driving and traffic signal data via Vehicle-to-Infrastructure (V2I) and Vehicle-to-Vehicle (V2V) communications. In the MPC-based control model, this paper mathematically formulates location-based traffic control devices and road geometry constraints using the geographic information from High-Definition (HD) maps. The location-based traffic control devices and road geometry constraints have the potential to improve the safety, energy, efficiency, driving comfort, and robustness of connected and automated driving on real roads by considering interrupted flow facility locations and road geometry in the formulation. We predict a set of uncertain driving states for the preceding vehicles through an online learning-based driving dynamics prediction model. We then solve a constrained finite-horizon optimal control problem with the predicted driving states to obtain a set of Eco-driving references for the controlled vehicle. To obtain the optimal acceleration or deceleration commands for the controlled vehicle with the set of Eco-driving references, we formulate a Distributionally Robust Stochastic Optimization (DRSO) model (i.e., a special case of data-driven optimization models under moment bounds) with Distributionally Robust Chance Constraints (DRCC) with location-based traffic control devices and road geometry constraints. We design experiments to demonstrate the proposed model under different traffic conditions using real-world connected vehicle trajectory data and Signal Phasing and Timing (SPaT) data on a coordinated arterial with six actuated intersections on Fuller Road in Ann Arbor, Michigan from the Safety Pilot Model Deployment (SPMD) project.

Fusing crash data and surrogate safety measures for safety assessment: Development of a structural equation model with conditional autoregressive spatial effect and random parameters

Most existing efforts to assess safety performance require sufficient crash data, which generally takes a few years to collect and suffers from certain limitations (such as long data collection time, under-reporting issue and so on). Alternatively, the surrogate safety measure (SSMs) based approach that can assess traffic safety by capturing the more frequent “near-crash” situations have been developed, but it is criticized for the potential sampling and measurement errors. This study proposes a new safety performance measure—Risk Status (RS), by fusing crash data and SSMs. Real-world connected vehicle data collected in the Safety Pilot Model Deployment (SPMD) project in Ann Arbor, Michigan is used to extract SSMs. With RS treated as a latent variable, a structural equation model with conditional autoregressive spatial effect and corridor-level random parameters is developed to model the interrelationship among RS, crash frequency, risk identified by SSMs, and contributing factors. The modeling results confirm the proposed interrelationship and the necessity to account for both spatial autocorrelation and unobserved heterogeneity. RS can integrate both crash frequency and SSMs together while controlling for observed and unobserved factors. RS is found to be a more reliable criterion for safety assessment in an implementation case of hotspot identification.

Classifying travelers' driving style using basic safety messages generated by connected vehicles: Application of unsupervised machine learning

Driving style can substantially impact mobility, safety, energy consumption, and vehicle emissions. While a range of methods has been used in the past for driving style classification, the emergence of connected vehicles equipped with communication devices provides a new opportunity to classify driving style using high-resolution (10 Hz) microscopic real-world data. In this study, location-based big data and machine learning are used to classify driving styles ranging from aggressive to calm. This classification can be used to customize driver assistance systems, assess mobility, crash risk, fuel consumption, and emissions. This study’s main objective is to develop a framework that harnesses Basic Safety Messages (BSMs) generated by connected vehicles to quantify instantaneous driving behavior and classify driving styles in different spatial contexts using unsupervised machine learning methods. To this end, a subset of the Safety Pilot Model Deployment (SPMD) with more than 27 million BSM observations generated by more than 1300 individuals making trips on diverse roadways and through several neighborhoods in Ann Arbor, Michigan, were processed and analyzed. To quantify driving style, the concept of temporal driving volatility, as a surrogate safety measure of unsafe driving behavior, was utilized and applied to vehicle kinematics, i.e., observed speeds and longitudinal/lateral accelerations. Specifically, six volatility measures are extracted and used for classifying drivers. K-means and K-medoids methods are applied for grouping drivers in aggressive, normal, and calm clusters. Clustering results indicate that not only does driving style vary among drivers, but the thresholds for aggressive and calm driving vary across different roadway types due to variations in environment and road conditions. The proportion of aggressive driving styles was also higher on commercial streets than on highways and residential streets. Notably, we propose a Driving Score to measure driving performance consistently across drivers.

Review of Usage of Real-World Connected Vehicle Data

GPS loggers and cameras aboard connected vehicles can produce vast amounts of data. Analysts can mine such data to decipher patterns in vehicle trajectories and driver–vehicle interactions. Ability to process such large-scale data in real time can inform strategies to reduce crashes, improve traffic flow, enhance system operational efficiencies, and reduce environmental impacts. However, connected vehicle technologies are in the very early phases of deployment. Therefore, related datasets are extremely scarce, and the utility of such emerging datasets is largely unknown. This paper provides a comprehensive review of studies that used large-scale connected vehicle data from the United States Department of Transportation Connected Vehicle Safety Pilot Model Deployment program. It is the first and only such dataset available to the public. The data contains real-world information about the operation of connected vehicles that organizations are testing. The paper provides a summary of the available datasets and their organization, and the overall structure and other characteristics of the data captured during pilot deployments. Usage of the data is then classified into three categories: driving pattern identification, development of surrogate safety measures, and improvements in the operation of signalized intersections. Finally, some limitations experienced with the existing datasets are identified.

Understanding speeding behavior from naturalistic driving data: Applying classification based association rule mining

Speeding is considered as one of the most significant contributing factors to severe traffic crashes. Understanding the associations between trip/driving/roadways features and speeding behavior is crucial for both researchers and practitioners. This research utilized naturalistic driving data collected by the Safety Pilot Model Deployment (SPMD) program and roadway features from a road inventory dataset - Highway Performance Monitoring System (HPMS), provided by the United States Department of Transportation (USDOT), to investigate the hidden rules that associated trip/driving/roadway features with speeding behavior. A classification-based association (CBA) algorithm was adopted to explore the hidden rules from two perspectives of speeding: speeding duration and speeding pattern. Results indicate that the combinations of longer trips (more than 60 min), driving on the roadways with a relatively higher functional class are highly associated with longer speeding events (speeding longer than 2 min). The moderate speeding events (speeding longer than 2 min and longer than 30 s) are found highly associated with the combination of driving on roadways with lower functional class, absence of a median and relatively short trip time (less than 30 min). The research also found the combinations of driving on roadways with relatively lower functional class, experienced congestion before a speeding event, and the presence of a median is a leading cause that triggers a higher speeding pattern (speeding more than 5mph above the speed limit). Furthermore, the moderate speeding pattern (speeding more than 1mph above the speed limit and less than 5mph of the speed limit) is associated with the combinations of factors like experiencing congestion before a speed event, driving on roadways with higher functional class and a relatively shorter trip (less than 30 min). The findings can help practitioners understand the composite effect of these factors more comprehensively and provide corresponding countermeasures to mitigate the negative consequences of speeding wherever possible. These can also help in calibrating driver behavior parameters for transportation-related simulation tools.

Big Data for Finite Population Inference: Applying Quasi-Random Approaches to Naturalistic Driving Data Using Bayesian Additive Regression Trees

Abstract Big Data are a “big challenge” for finite population inference. Lack of control over data-generating processes by researchers in the absence of a known random selection mechanism may lead to biased estimates. Further, larger sample sizes increase the relative contribution of selection bias to squared or absolute error. One approach to mitigate this issue is to treat Big Data as a random sample and estimate the pseudo-inclusion probabilities through a benchmark survey with a set of relevant auxiliary variables common to the Big Data. Since the true propensity model is usually unknown, and Big Data tend to be poor in such variables that fully govern the selection mechanism, the use of flexible non-parametric models seems to be essential. Traditionally, a weighted logistic model is recommended to account for the sampling weights in the benchmark survey when estimating the propensity scores. However, handling weights is a hurdle when seeking a broader range of predictive methods. To further protect against model misspecification, we propose using an alternative pseudo-weighting approach that allows us to fit more flexible modern predictive tools such as Bayesian Additive Regression Trees (BART), which automatically detect non-linear associations as well as high-order interactions. In addition, the posterior predictive distribution generated by BART makes it easier to quantify the uncertainty due to pseudo-weighting. Our simulation findings reveal further reduction in bias by our approach compared with conventional propensity adjustment method when the true model is unknown. Finally, we apply our method to the naturalistic driving data from the Safety Pilot Model Deployment using the National Household Travel Survey as a benchmark.

Integrating safety and mobility for pathfinding using big data generated by connected vehicles

With the emergence of the internet of things, pathfinding problems have recently received a significant amount of attention. Various commercial applications provide automated routing by considering travel time, travel distance, fuel consumption, complexity of the road, etc. However, many of these prospective applications do not consider route safety. Emergence of high-resolution big data generated by connected vehicles (CV) helps us to integrate safety into routing problem. The goal of this study is to address safety aspects in pathfinding problems by developing a methodological framework that simultaneously considers safety and mobility. To reach this goal, the concept of volatility is utilized as a surrogate safety performance measure to quantify route safety and driver behavior. The proposed framework uses CV big data and real-time traffic data to calculate safety indices and travel times. Measured safety indices include 5-year crash history, route speed and acceleration volatility, and driver volatility. Travel time and safety shape a cost function called “route impedance.” The algorithm has the flexibility for the user to predefine the weight for safety consideration. It also uses driver volatility to automatically increase safety weight for volatile drivers. To illustrate the algorithm, a numerical example is provided using an origin-destination pair in Ann Arbor, MI, and more than 42 million CV observations from around 2,500 CVs from the Safety Pilot Model Deployment (SPMD) were analyzed. The sensitivity analysis is performed to discuss the impact of penetration rate of CVs and time of the trip on the results. Finally, this paper shows suggested routes for multiple scenarios to demonstrate the outcome of the study. The results revealed that the algorithm might suggest different routes when considering safety indices and not just travel time.