No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Classifying travelers' driving style using basic safety messages generated by connected vehicles: Application of unsupervised machine learning
Abstract
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.
... Employing driving simulators is a viable and widely used option to ensure consistent environmental conditions for each study participant. Moreover, with assured repeatability, there is the flexibility to systematically modify elements like the road, traffic, weather, and other factors in a simple and replicable manner with low inherent risks and reduced costs [71], [102], [103]. There is substantiated evidence that a driving simulator is a valid tool for analyzing driving behavior, as there is a good agreement up to absolute validity between the behavior in a driving simulator and real-world driving [104], [105], This article has been accepted for inclusion in a future issue of this journal. ...
... The first factor represents four different driving styles: passive, rail, sportive, and the participants' trajectories replay. Driving style analysis is commonly treated as classification, and related work frequently distinguishes three classes [103], [114], [145], [146], [147]. As the second factor, all driving styles were evaluated in a randomized order in the two weather conditions: clear and rainy. ...
... The set of parameters, summarized in Table II, are denoted as the three variants: Passive, Rail, and Sportive for better readability. Such subjective classification is commonly used also in the literature [21], [91], [103], [175]. Figure 3 exemplarily illustrates the resulting different driving behaviors, together with the distribution of the absolute distances to the lane center for all curves. ...
Technological advancements focus on developing comfortable and acceptable driving characteristics in autonomous vehicles. Present driving functions predominantly possess predefined parameters, and there is no universally accepted driving style for autonomous vehicles. While driving may be technically safe and the likelihood of road accidents is reduced, passengers may still feel insecure due to a mismatch in driving styles between the human and the autonomous system. Incorporating driving style preferences into automated vehicles enhances acceptance, reduces uncertainty, and poses the opportunity to expedite their adoption. Despite the increased research focus on driving styles, there remains a need for comprehensive studies investigating how variations in the driving context impact the assessment of automated driving functions. Therefore, this work evaluates lateral driving style preferences for autonomous vehicles on rural roads, considering different weather and traffic situations. A controlled study was conducted with a variety of German participants utilizing a high-fidelity driving simulator. The participants experienced four different driving styles, including mimicking of their own driving behavior under two weather conditions. A notable preference for a more passive driving style became evident based on statistical analyses of participants’ responses during and after the drives. This study could not confirm the hypothesis that people prefer to be driven by mimicking their own driving behavior. Furthermore, the study illustrated that weather conditions and oncoming traffic substantially influence the perceived comfort during autonomous rides. The gathered dataset is openly accessible at https://www.kaggle.com/datasets/jhaselberger/idcld-subject-study-on-driving-style-preferences .
... To differentiate drivers, most existing driver profiling methods rely on kinematic data on the premise that all the drivers drive on the same route [3,[12][13][14] or the same type of roads [15,16] (see details in Sect. 2). ...
... Finally, the behaviors of the three groups associated with each road type were characterized as moderate, unstable, or cautious, respectively. Similarly, Mohammadnazar et al. [16] profiled hundreds of drivers, who drove on highways, residential streets, and commercial streets. These drivers were characterized as aggressive, normal, and calm for each road type, respectively. ...
... To fairly compare drivers' driving behaviors, we need to make sure that the comparison is only performed when drivers have a similar driving environment. Intuitively, we assume similar roads constitute a similar driving environment for the drivers, which follows the assumptions made by the prior studies [15,16]. Then, prior to comparing the drivers, we measure the similarity among different roads and determine which ones are deemed as the same types. ...
Driver profiling is a widely used tool in fleet management and driver-specific insurance because it differentiates drivers based on their driving behaviors, such as aggressive and non-aggressive, which correspond to different levels of driving risk. However, most existing driver profiling methods require all drivers to drive on the same predefined route or type of roads, simply to make sure their driving behaviors are comparable. This premise makes these methods not be able to profile drivers who drive on arbitrary roads, which constitute the real-world scenarios for most drivers. To enable the profiling of drivers using their naturalistic driving data, i.e., driving trajectories recorded while they were driving on arbitrary roads at their own free will, in this paper, we propose a novel method named cLustering rOads And Drivers Successively (LOADS). Specifically, LOADS first categorizes the roads into different types using the extracted characteristics of all drivers driving on the respective roads. It then groups drivers into different clusters to obtain their profile labels (e.g., aggressive or non-aggressive) using the extracted driving characteristics on each road type. We conduct extensive experiments using two real-world driving trajectory datasets comprising thousands of driving trajectories of hundreds of drivers. Statistical analysis results indicate that the driver groups identified by LOADS have significantly different driving styles. To the best of our knowledge, LOADS is the first method that focuses on profiling drivers who drive on arbitrary roads, showing a great potential to enable real-world driver profiling applications.
... Unsupervised learning approaches can be used to classify by analyzing the inherent relationships in driving behavior data without labeled data and can be categorized into the following approaches: (1) Principal Components Analysis (PCA) and Hierarchical Clustering Analysis (HCA) (Chada et al. 2022;Constantinescu, Marinoiu, and Vladoiu 2010), (2) Gaussian Mixed Model (Song et al. 2023), (3) k-means clustering (de Zepeda et al. 2021;Gao et al. 2020;Ma et al. 2021;Mohammadnazar, Arvin, and Khattak 2021), and others (Marina Martinez et al. 2018). For example, Constantinescu, Marinoiu, and Vladoiu (2010) categorized driving styles based on risk tendencies and used PCA and HCA to identify driving behaviors. ...
... Driving behaviors were modeled by PCA and HCA using collected driving data. Mohammadnazar, Arvin, and Khattak (2021) developed a framework for quantifying instantaneous driving behaviors. Driving styles were clustered by the K-means approach based on basic safety messages. ...
... Various methods have been used to investigate driving behavior. According to Jeihani and Banerjee (2018) and Mohammadnazar et al. (2021), data collection methods include surveys, questionnaires, simulations, roadside camera observations, and naturalistic experiments. Using data that have been extracted from naturalistic experiments is suitable due to the high reliability. ...
... Initially, these centroid points are assigned randomly and evolve through each iteration. The k-means clustering algorithm is a well-established technique in cluster analysis, and is renowned for its efficacy in various applications (Khanfar et al. 2022b;Mohammadnazar et al. 2021). ...
Driving behavior is a unique driving habit of each driver, and it has a significant impact on road safety. Classifying driving behavior and introducing policies based on the results can reduce the severity of crashes on the road. Roundabouts are particularly interesting because of the interconnected interaction between different road users at the roundabouts, in which different driving behavior is hypothesized. This study investigated driving behavior at roundabouts in a mixed traffic environment using data-driven unsupervised machine learning to classify driving behavior using a data set from three roundabouts in Germany. We used a data set of vehicle kinematics for a group of different vehicles and vulnerable road users (VRUs) at roundabouts and classified them into three categories (i.e., conservative, normal, and aggressive). The results showed that most drivers proceeding through a roundabout can be classified into two driving styles—conservative, and normal—because traffic speeds in roundabouts are relatively lower than at other signalized and unsignalized intersections. The results
also showed that about 77% of drivers who interacted with pedestrians or cyclists were classified as conservative drivers, compared with about 42% of drivers who did not interact with pedestrians or cyclists, and about 51% of all drivers. Drivers tend to behave abnormally when they interact with VRUs at roundabouts, which increases the risk of crashes when an intersection is multimodal. The results of this study could help to improve the safety of roads by allowing policymakers to determine effective and suitable safety countermeasures. The results also will be beneficial for advanced driver-assistance systems (ADAS) as the technology is deployed in a mixed traffic environment.
... Analysing driving behaviour numerically can be challenging due to the multitude of variables that influence it. As a result, there are two primary categories of studies: one focusing predominantly on survey results and the other on vehicle information [2]. Quality data are essential for risk assessment, but obtaining high-quality pre-accident data is extremely challenging. ...
... The clustering analysis aims to determine the optimal number of clusters for distinguishing normal and risky driving profiles. The clustering process was applied to risky driving profiles [2]. ...
Advances in real‐time basic safety message (BSM) data from sensor‐equipped vehicles have created new opportunities for driving risk assessments. This paper presents a machine learning approach using BSM data to provide fine‐grained risk assessments, focusing on safety‐critical events (SCEs) related to driving profiles, vehicle states, and road conditions. This approach formulates a bi‐level risk indicator: one level measures the observable frequency of SCEs, while the other estimates their likelihood. The coarse level calculates risk scores by classifying driving profiles as normal or risky based on SCE frequency. The fine level refines these scores by comparing normal and risky profiles using key features from a feature learning model. This combined system accounts for recent driving behaviours and road/weather conditions within a vehicle‐to‐everything (V2X) environment, addressing high data dimensionality and imbalance. A comprehensive case study using 1 year of data from pilot V2X infrastructure in Tampa, Florida, demonstrates the efficacy of this approach, showing practical applications of the SCE‐based risk indicator and combinatorial feature learning while also highlighting the real‐world utility of the assessment method in providing a detailed and actionable view of driving risk based on V2X information.
... Extending it to full 3D space with altitude control will enable application to more complex aerial missions. Third, future work could explore adaptive strategies inspired by driving behavior research [36,37]. Such techniques could provide enhancement for UAV formations to handle multi-agent interactions, uncertainty, and individual differences better. ...
Formation flying of multiple unmanned aerial vehicles (UAVs) has attracted much attention for its versatility in cooperative tasks. In this paper, a distributed formation planning method is proposed for UAVs. First, we design a path searching algorithm, swarm-A*, which can enhance the cohesion of a swarm, i.e., preventing the disintegration of the swarm when it encounters an obstacle. Then, after waypoint reallocation, a formation trajectory optimization framework is formulated. Smooth formation trajectories for UAVs to travel safely in obstacle-laden environments can be obtained by solving the optimization problem. Next, a tracking controller based on sliding mode control is designed, ensuring that the UAVs follow the planned formation trajectories under dynamic constraints. Finally, numerical simulations and experiments are conducted to validate the effectiveness of the proposed method.
... Driving behavior encompasses various variables and factors, including driving performance, environmental awareness, risk-taking propensity, and reasoning abilities (5). Abnormal driving behavior refers to reckless actions that deviate from safe and normal driving, posing risks to the driver, passengers, and other road users, and typically occurs within a short period of time (6). ...
Detecting abnormal driving behavior is critical for road traffic safety and the evaluation of drivers’ behavior. With the advancement of machine learning (ML) algorithms and the accumulation of naturalistic driving data, many ML models have been adopted for abnormal driving behavior detection (also referred to in this paper as “anomalies”). Most existing ML-based detectors rely on (fully) supervised ML methods, which require substantial labeled data. However, ground truth labels are not always available in the real world, and labeling large amounts of data is tedious. Thus, there is a need to explore unsupervised or semi-supervised methods to make the anomaly detection process more feasible and efficient. To fill this research gap, this study analyzes large-scale real-world data revealing several abnormal driving behaviors (e.g., sudden acceleration, rapid lane-changing) and develops a hierarchical extreme learning machine (HELM)-based semi-supervised ML method using partly labeled data to accurately detect the identified abnormal driving behaviors. Moreover, previous ML-based approaches predominantly utilized basic vehicle motion features (such as velocity and acceleration) to label and detect abnormal driving behaviors, while this study seeks to introduce event-level safety indicators as input features for ML models to improve the detection performance. Results from extensive experiments demonstrate the effectiveness of the proposed semi-supervised ML model with the introduced safety indicators serving as important features. The proposed semi-supervised ML method outperforms other baseline semi-supervised or unsupervised methods as far as various metrics are concerned: for example, it delivers the best accuracy at 99.58% and the best F1-score at 0.9913. The ablation study further highlights the significance of safety indicators for advancing the detection performance of abnormal driving behaviors.
... However, unscaled precision measures are often difficult to compare across studies. Scaled precision metrics, like the coefficient of variation (CV), recommended by Mohammadnazar et al. [28] for vehicle mobility prediction, face challenges in train delay prediction. This limitation arises from frequent near-zero delays, causing the CV to be highly sensitive to minor mean fluctuations, particularly when the mean of train delays in the denominator is zero or negative, misleading the CV results. ...
The surging demand for Intelligent Transportation Systems (ITS) to deliver advanced train-related Information for dispatchers and passengers has spurred the development of advanced train delay prediction models. Despite considerable efforts devoted to developing methodologies that can be used to model train operation conditions and produce anticipated train delays, the evaluation strategies for train delay prediction models remain under-researched, particularly evident when accuracy is always found to be the only determinant in model selection. The absence of a standardised evaluation procedure for assessing the effectiveness of these prediction models has hindered the practical implementation of these models. To bridge this gap, the study conducted a systematic literature review on data-driven train delay prediction models and introduced the novel AP-GRIP (Accuracy, Precision, Generalisability, Robustness, Interpretability, Practicality) evaluation framework. The framework covers six key aspects across overall, spatial, temporal, and train-specific dimensions, providing a systematic approach for the comprehensive assessment of train delay prediction models. Each aspect and dimension is thoroughly discussed and synthesised with its definitions, measuring metrics, and important considerations. A critical discussion clarifies several interactions, such as predetermined objectives, desired outputs, model type, benchmark models, and data availability, resulting in a logical framework for assessing train delay prediction models. The proposed framework uncovers inadequate prediction patterns, offering insights on when, where, and why the prediction models excel and fall short, assisting end-users in determining model suitability for specific prediction tasks.
... A data source unique to connected vehicles is Basic Safety Messages (BSMs), which are exchanged between connected vehicles using onboard units (OBUs). Several studies have discussed how these messages can provide pertinent vehicle information (Abdelkader et al., 2021;Dokur & Katkoori, 2022;Kim & Kim, 2020;Miucic, 2019;Mohammadnazar et al., 2021;Wang et al., 2021). The BSM data set was used to capture variations in vehicle control (Arvin et al., 2019). ...
With over six million annual police-reported crashes in the United
States, surveys on law enforcement’s involvement in crash investigations
are rare. A unique and innovative survey is designed in
the study to address the contemporary issues of vehicle automation,
focusing on the integration of Connected and Automated
Vehicles (CAVs) data to advance crash investigation accuracy.
Utilizing both descriptive statistics and rigorous factor analysis, the
study surveyed 61 Tennessee law enforcement officers with crash
investigation duties. The survey included respondents’ vehicle
crash investigation experience, training exposure, and familiarity
with automated vehicle technologies. The study’s results indicate
that officers have experience utilizing video camera footage in
crash investigations. The findings revealed law enforcement officers’
moderate awareness of LiDAR and limited understanding of
millimeter-wave radar and ultrasound sensors. Survey respondents
acknowledged that access to vehicle trajectory data from CAV sensors
could significantly aid crash investigations. The survey
assessed the law enforcement needs for standardized CAV data
retrieval and identified critical training areas, emphasizing a comprehensive
grasp of CAV technologies and data processes. The
study establishes the foundation for a customized training program,
emphasizing the critical need for law enforcement to keep
pace with rapidly advancing vehicular technologies in the realm
of crash investigation.
... Research has highlighted the critical role of dynamically adjusting control parameters based on road conditions and driver preferences to improve the adaptability and performance of autonomous systems. Some studies have concentrated on utilizing machine learning and predictive modeling to understand and classify driving styles [27,28], while others have focused on integrating social and individual preferences into decision-making processes for specific scenarios, such as lane changes or urban intersections [29,30]. Furthermore, methodologies inspired by human decision-making frameworks have been proposed to address the complexities of mixed-traffic environments [31]. ...
This study introduces a motion-sickness-reducing control strategy aimed at enhancing ride comfort in Electric Autonomous Vehicles (EAVs). For lateral control, the forward look-ahead distance was adaptively adjusted based on the Motion Sickness Dose Value (MSDV) analysis from ISO 2631-1, effectively mitigating lateral acceleration and its motion-sickness-related frequency components, leading to a reduced MSDV. For longitudinal control, Linear Quadratic Regulator (LQR) optimal control was applied to minimize acceleration, complemented by a band-stop filter specifically designed to attenuate motion-sickness-inducing frequencies in the acceleration input. The bandwidth of the band-stop filter used in this study was designed based on the motion-sickness frequency weighting specified in ISO 2631-1. The simulation results of the proposed control indicate a significant reduction in MSDV, decreasing from 16.3 to 10.46, achieving up to a 35.8% improvement compared to comparative control methods. While the average lateral position error was slightly higher than that of the comparative controller, the vehicle consistently maintained lane adherence throughout path-following tasks. These findings underscore the potential of the proposed method to simultaneously mitigate motion sickness and achieve a robust path-following performance in autonomous vehicles.
... Taking into consideration the challenges in acquiring driver style classifications in real application settings, current studies mainly employ unsupervised learning algorithms to detect and categorize driving styles. Random forest decision methods, support vector machines, and artificial neural networks are often the most utilised methods for the categorization of driving style [13,19,22,23]. In contrast to the literature, the study presented in this article concerns smartphone data collected on the passenger side. ...
This paper presents a machine learning analysis of driving behaviors in informal minibus taxis, focusing on both controlled and uncontrolled environments. Informal minibus taxis play a crucial role in urban transportation, particularly in developing countries, yet their driving patterns and safety implications remain under-explored. We utilize exploratory factor analysis to analyze data collected from smartphone GPS carried by a passenger of a minibus taxi, identifying key driving behaviors and patterns. Our study highlights significant differences in driving styles between controlled and uncontrolled environments, offering insights into safety and efficiency. The findings provide valuable information for policymakers, transportation planners, and technology developers aiming to enhance urban mobility and safety in the informal transport sector.
... Besides this, Cluster Analysis -supported by unsupervised machine learning [26] is used in countless research fields as a general tool to automatically define classifications within a heterogeneous set of elements based on data similarity. This has also produced novel application modalities of this method in mobility itself: some examples include the classification of travellers' driving style [27] or the evaluation of street safety related to weather [28]. ...
... In car-following scenarios, parameters such as time to collision (TTC) [4,5], overspeed frequency [6], extreme value of steering wheel angle change rate [7], and speed and acceleration of the preceding vehicle [2] are usually selected. In lane-changing scenarios, parameters such as lanechanging time headway [6], maximum steering angle during lane-changing [8], lateral acceleration [9], lane-changing time, and lane-changing interval allowed for vehicles [10] are usually selected. When selecting features, parameters such as speed fluctuation, throttle activity index, and engine load [11] could be considered to combine driving style with energy consumption [12]. ...
At present, achieving efficient, sustainable, and safe transportation has led to increasing attention on driving behavior recognition and advancements in autonomous driving. Identifying diverse driving styles and corresponding types is crucial for providing targeted training and assistance to drivers, enhancing safety awareness, optimizing driving costs, and improving autonomous driving systems responses. However, current studies mainly focus on specific driving scenarios, such as free driving, car-following, and lane-changing, lacking a comprehensive and systematic framework to identify the diverse driving styles. This study proposes a novel, data-driven approach to driving-style recognition utilizing naturalistic driving data NGSIM. Specifically, the NGSIM dataset is employed to categorize car-following and lane-changing groups according to driving-state extraction conditions. Then, characteristic parameters that fully represent driving styles are optimized through correlation analysis and principal component analysis for dimensionality reduction. The K-means clustering algorithm is applied to categorize the car-following and lane-changing groups into three driving styles: conservative, moderate, and radical. Based on the clustering results, a comprehensive evaluation of the driving styles is conducted. Finally, a comparative evaluation of SVM, Random Forest, and KNN recognition indicates the superiority of the SVM algorithm and highlights the effectiveness of dimensionality reduction in optimizing characteristic parameters. The proposed method achieves over 97% accuracy in identifying car-following and lane-changing behaviors, confirming that the approach based on naturalistic driving data can effectively and intelligently recognize driving styles.
... This is to identify the behavior of drivers as a significant safety parameter. Higher volatility measures were also observed to indicate a greater likelihood of the driver being unstable and risky, suggesting increased aggressiveness [29,30]. We used ten different volatility measures ( VM i ), as shown in Table 1. ...
This study is a significant endeavor involving the development and testing of a comprehensive methodology to incorporate driving behavior into the analysis and prediction of vehicle fuel consumption. It underscores the crucial importance of understanding how different driving behavior affect fuel efficiency. The framework we present is a theoretical construct and a practical tool. It provides a robust, multi-step process for linking driving behavior to fuel consumption, leveraging both traditional statistical methods and advanced machine learning techniques to derive actionable insights. To test the framework, we used a naturalistic data that includes about 5408 different road users in a mixed traffic environment and urban settings in Germany. We applied a microscopic fuel consumption model to calibrate the framework and an unsupervised clustering algorithm to classify the behavior of the driver interacting with each other and with vulnerable road users. The framework includes developing Linear regression model as a baseline, which yields an R-squared of 0.511 and a Mean Squared Error (MSE) of 0.031, indicating moderate predictive accuracy. The final step includes choosing Random Forest as a better model, which achieves a higher R-squared of 0.956 and a lower MSE of 0.003. We also found that conservative and aggressive driving leads to significantly higher and more discrepancy in fuel consumption than normal driving behavior. These insights can promote more efficient driving practices, leading to significant fuel savings and environmental benefits.
... These vehicles have the potential to blur the boundaries between public and private transportation services, making MaaS appear as a viable compromise [59]. In particular, various AI-based models have been implemented in different MaaS solutions, as in classifying drivers' driving styles [60] and to develop optimization models to manage bike-sharing fleets in smart cities [61][62][63]. Furthermore, unlike previous studies where the impact of such solutions was not directly perceived by users, Rajabi et al. [64] introduced an innovative AI approach that ensures that MaaS service users receive a personalized itinerary, capable of adapting to individual modal preferences [65,66] and weather conditions. ...
City planners are facing an emergency to develop, promote, and implement actions that allow the mobility needs of people and businesses in cities (and their surroundings) to be satisfied to assure a better quality of life. Among the different actions to promote, there is public transport, which should be the cornerstone of sustainable urban mobility. The only way to achieve the transition from private to public is by integrating services into a multimodal network and then encouraging interchange between different modes of transportation. In this context, the development of mobility as a service (MaaS) solutions is significant, and different studies have been developed in recent years. The paper thus introduces a scientometric review of such a topic in relation to the Sustainable Development Goals of Agenda 2030. The study focuses on the current state of MaaS implementation, trends, and research gaps, with an in-depth analysis of emerging themes, based on 819 documents selected from the WoS and Scopus databases. Introducing the database, and research methodology, an accurate interpretation of the data generated by the bibliometric analysis, and the primary evaluation parameters are outlined. The identified studies have been then categorized into three thematic groups with the intention of offering a comprehensive study that identifies the shortcomings and difficulties in the research carried out in these areas up to today. Particular attention is paid to how this research relates to the Sustainable Development Goals.
... Some usual tasks performed by AI algorithms are classification, regression, clustering, or pattern recognition within a large dataset [41]. Currently, ML is utilized for many different things, for example in healthcare field [42][43][44][45], in Soil and Mineral analysis [46,47], in analyzing driving styles [48], and in handwritten recognition [41] etc., including recommender systems, email filtering, Facebook auto-tagging, image identification, and speech recognition [49]. ML is broadly divided into two categories: supervised ML and unsupervised ML [50]. ...
This paper addresses the challenge of objectively determining a website’s personality by developing a methodology based on automated quantitative analysis, thus avoiding the biases inherent in human surveys. Utilizing a database of 3000 websites, data extraction tools gather relevant data, which are then analyzed using Artificial Intelligence (AI) techniques, including machine learning (ML) and natural language processing. Four ML algorithms—K-means, Expectation Maximization, Hierarchical Agglomerative Clustering, and DBSCAN—are implemented to assess and classify website personality traits. Each algorithm’s strengths and weaknesses are evaluated in terms of data organization, cluster flexibility, and handling of outliers. A software tool is developed to facilitate the research process, from database creation and data extraction to ML application and results analysis. Experimental validation, conducted with identical training and testing datasets, achieves a success rate of up to 94% (with an Error of ≤50%) in accurately identifying website personality, which is validated by subsequent surveys. The research highlights significant relationships between website attributes and personality traits, offering practical applications for website developers. For instance, developers can use these insights to design websites that align with business goals, enhance customer engagement, and foster brand loyalty. Additionally, the methodology can be applied to creating culturally resonant websites, thus supporting New Zealand’s cultural initiatives and promoting cross-cultural understanding. This research lays the groundwork for future studies and has broad applicability across various domains, demonstrating the potential for automated, unbiased website personality classification.
... A data source unique to connected vehicles is Basic Safety Messages (BSMs), which are exchanged between connected vehicles using onboard units (OBUs). Several studies have discussed how these messages can provide pertinent vehicle information (Abdelkader et al., 2021;Dokur & Katkoori, 2022;Kim & Kim, 2020;Miucic, 2019;Mohammadnazar et al., 2021;Wang et al., 2021). The BSM data set was used to capture variations in vehicle control (Arvin et al., 2019). ...
Some of the 6 million plus police-reported crashes in the US require detailed investigations. As connected and automated vehicles (CAVs) gain popularity and are also involved in crashes, new data from CAVs can be used to improve the accuracy of crash investigations. In addition to information from Event Data Recorders, CAV data on vehicle trajectories and pre-crash conditions can enhance the understanding of causation factors. This study investigates how CAV data can advance crash investigations by answering research questions about what pertinent information is lacking in crash investigations and how law enforcement can use CAV data. This study addresses these questions by performing a survey of 61 law enforcement officials working in crash investigations in the state of Tennessee. The survey included respondents’ vehicle crash investigation experience, training exposure, and familiarity with automated vehicle technologies. The survey showed that respondents have previous experience working with video camera footage in crash investigations, and responses indicate that camera footage would provide the most helpful new information when investigating a crash. The survey also revealed law enforcement officials’ moderate familiarity with lidar and low familiarity with millimeter-wave radar and ultrasound sensors. Survey respondents also indicated that data such as vehicle trajectories made available through automated vehicle sensors would be useful during crash investigations. The survey revealed a need for standardization in CAV data retrieval and training processes. It resulted in a list of pertinent training topics for law enforcement that prioritizes a thorough understanding of CAV technology and data retrieval processes.
... By doing this, each driver has their personalised IDM model parameters meanwhile following a particular driving style. It is assumed that drivers maintain a consistent driving style throughout the simulation period (Mohammadnazar, Arvin, and Khattak 2021). Each traffic scenario is simulated with 100 seeds to enhance the reliability of the results. ...
... Our methodology leverages advanced machine learning models, including classification algorithms and neural networks, to adapt educational content to the diverse needs of a cohort of 450 university students. Machine learning has been chosen for its ability to handle large volumes of data and its potential to reveal complex patterns that may not be immediately evident [4], [5]. These models can dynamically adjust educational materials, providing personalization and adaptability previously unattainable with traditional methods. ...
Given the inherent diversity in learning styles and rhythms, the current educational landscape demands continuous adaptation toward methodologies that enhance individualized learning. This study addresses the effectiveness of learning personalization using machine learning models to adapt educational content to individual learning styles. Focusing our attention on a cohort of 450 university students, we implemented classification algorithms and neural networks to diagnose learning styles and personalize educational resources accordingly. The results are revealing: the students’ average grades experienced a significant increase, going from 70 to 75 points on a scale of 100 after the personalized intervention. Additionally, increased engagement was recorded, evidenced by more substantial interaction with educational materials tailored to their learning preferences. These findings suggest that personalization of learning is a powerful and effective tool that can improve both academic performance and students’ educational experience. This work confirms the relevance of educational personalization supported by artificial intelligence and provides a practical model for its effective implementation. The implications of this study are particularly pertinent to the evolution of pedagogical practices and curriculum design in the digital age.
... In addition to these dynamics-oriented indicators, the model parameters of classical mathematical driving behavior models are also predicted [84], [93]- [95]. Moreover, scores, such as sportiness or aggressiveness, are derived using predefined calculation procedures [44], [89], [96], [97]. In contrast to the broader driving style classes, the objective indicators of driving behavior offer the advantage of being directly integrable into the personalization of driver assistance systems or automated driving functions through constraints or target variables. ...
There is evidence that the driving style of an autonomous vehicle is important to increase the acceptance and trust of the passengers. The driving situation has been found to have a significant influence on human driving behavior. However, current driving style models only partially incorporate driving environment information, limiting the alignment between an agent and the given situation. Therefore, we propose a situationaware driving style model based on different visual feature encoders pretrained on fleet data, as well as driving behavior predictors, which are adapted to the driving style of a specific driver. Our experiments show that the proposed method outperforms all evaluated baselines significantly and forms plausible situation clusters. Furthermore, we found that feature encoders pretrained on our dataset lead to more precise driving behavior modeling. In contrast, feature encoders pretrained supervised and unsupervised on different data sources lead to more specific situation clusters, which can be utilized to constrain and control the driving style adaptation for specific situations. Moreover, in a real-world setting, where driving style adaptation is happening iteratively, we found the MLP-based behavior predictors achieve good performance initially but suffer from catastrophic forgetting. In contrast, behavior predictors based on situationdependent statistics can learn iteratively from continuous data streams by design. Overall, our experiments show that important information for driving behavior prediction is contained within the visual feature encoder. The dataset is publicly available at huggingface.co/datasets/jHaselberger/SADC-SituationAwareness-for-Driver-Centric-Driving-Style-Adaptation.
... [50] K-means methodologies were applied to group drivers in aggressive, normal, and calm clusters to quantify driving style. [51] In addition, DL was applied to monitor the road surface. The results revealed that the classification accuracy of concrete panels, cobblestones, asphalt, and dirt roads by a convolutional neural network (CNN) was up to 85%. ...
Collaborative perception between a vehicle and the road has the potential to enhance the limited perception capability of autonomous driving technologies. With this background, self‐powered vehicle‐road integrated electronics (SVRIE) with a multilevel fractal structure is designed to play a dual role, including a SVRIE device integrated into vehicle tires and a SVRIE array embedded into a road surface. The pressure sensing capability and anti‐crosstalk performance of the SVRIE array are characterized separately to validate the feasibility of applying the SVRIE in a cooperative vehicle‐infrastructure system. It is demonstrated that the SVRIE based on the multi‐layered fractal structure exhibits maximum performance in collaborative sensing and interaction between vehicles and road information, such as vehicle motion, road surface condition, and tire life cycle health monitoring. Traditional data analysis methods are often of questionable accuracy. Therefore, a convolutional neural network is used to classify the vehicle and road conditions with accuracy of at least 88.3%. The transfer learning model is constructed to enhance the road surface identification capabilities with 100% accuracy. The accuracies of the vehicle tire motion recognition and tire health monitoring are 97% and 99%, respectively. This work provides new ideas for collaborative perception between vehicles and roadsides.
... The mean and standard deviation of the features in the clusters were used to determine the driving styles they represented. Since features such as speed and acceleration positively correlate with risky driving (Arvin et al. 2019;Mohammadnazar et al. 2021), clusters 1, 2, and 3 correspond to aggressive, normal, and cautious driving style, respectively. ...
Objectives:
Aggressive driving behavior can lead to potential traffic collision risks, and abnormal weather conditions can exacerbate this behavior. This study aims to develop recognition models for aggressive driving under various climate conditions, addressing the challenge of collecting sufficient data in abnormal weather.
Methods:
Driving data was collected in a virtual environment using a driving simulator under both normal and abnormal weather conditions. A model was trained on data from normal weather (source domain) and then transferred to foggy and rainy weather conditions (target domains) for retraining and fine-tuning. The K-means algorithm clustered driving behavior instances into three styles: aggressive, normal, and cautious. These clusters were used as labels for each instance in training a CNN model. The pre-trained CNN model was then transferred and fine-tuned for abnormal weather conditions.
Results:
The transferred models showed improved recognition performance, achieving an accuracy score of 0.81 in both foggy and rainy weather conditions. This surpassed the non-transferred models' accuracy scores of 0.72 and 0.69, respectively.
Conclusions:
The study demonstrates the significant application value of transfer learning in recognizing aggressive driving behaviors with limited data. It also highlights the feasibility of using this approach to address the challenges of driving behavior recognition under abnormal weather conditions.
... Driving style refers to the way the driver chooses to drive, which is an instant preference behavior or driving habit [32], especially how the driver puts pressure on the acceleration and brake pedals, which have a significant impact on the vehicle's speed. Under the same environmental factors and route, different driving styles will produce completely different speed profiles. ...
The estimation of energy consumption under real-world driving conditions is a prerequisite for optimizing bus scheduling and meeting the requirements of route operation, thereby promoting the large-scale application of battery electric buses. However, the limitation of data accuracy and the uncertainty of many factors, such as weather conditions, traffic conditions, and driving styles, etc. make accurate energy consumption estimation complicated. In response to these challenges, a new method for estimating the energy consumption of battery electric buses (BEBs) is proposed in this research. This method estimates the speed profiles of different driving styles and the energy consumption extremes using real-world driving data. First, this research provides the constraints on speed formed by environmental factors including weather conditions, route characteristics, and traffic characteristics. On this basis, there are two levels of estimation for energy consumption. The first level classifies different driving styles and constructs the corresponding speed profiles with the time interval (10 s), the same as real-world driving data. The second level further constructs the speed profiles with the time interval of 1 s by filling in the first-level speed profiles and estimating the energy consumption extremes. Finally, the estimated maximum and minimum value of energy consumption were compared with the true value and the results showed that the real energy consumption did not exceed the extremes we estimated, which proves the method we proposed is reasonable and useful. Therefore, this research can provide a theoretical foundation for the deployment of battery electric buses.
To achieve real-time detection on resource-constrained edge devices, a lightweight vehicle detection algorithm named YOLO-edge was developed based on the YOLOv5s framework. In YOLO-edge, a slim neck was designed to reduce the computational cost in the feature channel fusion process. A modified fast spatial pyramid pooling technique enhances feature extraction efficiency. A multi-scale feature fusion architecture with a broad receptive field integrates shallow and deep features, preserving details for comprehensive feature representation. The loss function was optimized to improve the accuracy of bounding box predictions. Experimental results demonstrate that YOLO-edge achieves a 5.1% increase in mean average precision (mAP) compared to YOLOv5s. Real-time detection rates of 34 and 47 frames per second (FPS) are achieved on edge devices Jetson TX2 and Jetson Orin NX, respectively. YOLO-edge exhibited superior speed and accuracy on edge devices, outperforming state-of-the-art YOLO methods in vehicle detection.
The realization of personalized lane-changing (LC) for intelligent vehicles (IVs) is important for enhancing the social acknowledgment, user acceptance, adaptability, and trust of IVs. The LC style classification of human drivers represents a crucial foundation for achieving personalized LC. Therefore, this study constructs an LC style classification method based on driving behavioral primitives, which enables the classified LC styles to fully embody the implicit behavioral semantics and patterns of human drivers. First, a disentangled sticky hierarchical Dirichlet process hidden Markov model is proposed for the LC behavioral segment segmentation. The model can suppress frequent transitions of the hidden states, and vector autoregression is used to accurately describe the LC explicit behavioral parameters. Subsequently, the K-shape is employed to cluster all LC behavior segments to obtain interpretable and reasonable LC behavior primitives. Then, clustering features based on the LC behavioral primitives are constructed. Finally, LC styles are classified using density peak clustering, which does not require a manual specification of the number of clustering centers. Verification is performed on the Next Generation Simulation dataset, and the results indicate that this method can accurately and reasonably classify LC styles. The quantitative comparison with four state-of-the-art methods further demonstrates the advantages of the proposed method in LC style classification and confirms the effectiveness of introducing LC behavioral primitives.
Understanding the indigenous driving styles of individuals is significant to avoid oversighting the behavioral generalization across varying geographical locations. The past research on driving style classification mostly focused on identifying the driving patterns using the kinematic feature magnitudes. The variation in the instantaneous driving decisions termed as “driving volatility” is not explored in the context of performance assessment. In this regard, the present study proposes a methodology to explore the driving styles of Indian drivers, using both the magnitude and variation exhibited in the short-term driving decisions. The real-time driving profiles of 47 professional car drivers were collected and segmented into maneuvers based on the respective driving regimes. The performance features representative of each maneuver are extracted, defining 12 measures of driving volatility. The K-means clustering was performed on the event dataset at two-levels, which resulted in four patterns of driving styles under acceleration and braking regimes. The results showed that, a driver can exhibit speedy and aggressive maneuvers in a stable as well as in a highly volatile pattern. The generated driving style profiles at individual-level highlight the behavioral changes in drivers pertained to external influencing factors, and helps to identify the aberrations performed in each trip.
In most developing countries, a significant percentage of households rely on public transport, specifically informal minibus taxis. In South Africa, this mode of transport is more regularly involved in road traffic accidents compared to other modes. However, very few studies have focused on analyzing driving in this transportation mode from the perception of the commuter. In this study, the analysis is carried out using qualitative (questionnaires) and quantitative (speed and acceleration) data with the aim of finding factors that characterize public transportation drivers, specifically understanding how minibus taxi drivers differ from other drivers and how the regulatory environment influences their on-road behavior. The personality and skills of the drivers are shown to be the two main factors to analyze. It is shown that minibus taxi drivers perform lower than ride-hailing drivers. In addition, their driving is more aggressive in a controlled environment, while it is more reckless in an uncontrolled environment. Cultural, training, and technology-oriented actions are suggested to improve the on-road driving of the minibus taxi in the chosen study area.
Vehicle-to-Everything (V2X) infrastructure generates a vast amount of data from sensor-equipped vehicles and road infrastructure. The availability of such data provides new opportunities to explore and understand drivers' behaviors in diverse scenarios and how road characteristics affect driving actions. Effectively extracting and communicating driving behaviors from this huge volume of data in a timely manner presents a significant challenge. In this paper, we introduce a novel approach that combines machine learning and visualization techniques to support the exploration and analysis of driving behaviors using extensive vehicle status data from V2X infrastructure. Our approach consists of three core components: a spatiotemporal driving behavior feature learning model, a combined individual- and road-level driving behavior clustering method, and a multi-level visualization framework. The feature learning model for driving behavior detection employs a bi-directional spatiotemporal attentive long short-term memory network mechanism to extract essential individualized driving behaviors. Time series vehicle status and location data are then transformed and represented as feature vectors so that an unsupervised clustering method can be applied to identify representative types of driving behaviors and assign clustering labels to individual feature vectors. Furthermore, the model aggregates clustering labels by road, enabling the examination of driving patterns in the context of road conditions. To facilitate the interpretation of feature vectors and clustering results, our visualization framework offers a set of multi-level designs that support visual analytics on both trajectory and road levels. This approach contributes to driving analysis by utilizing massive real-world V2X data. It specifically addresses existing limitations in combining machine learning with visualization to support intelligent discovery of traffic patterns in an automatic manner. This not only enhances efficiency, but also ensures improved explainability and transparency for interpreting results.
Traffic accidents, mostly caused by human errors, pose a significant challenge in automotive safety. This paper introduces TUN-DAS, a novel Time-series analysis and Unsupervised learning-based Driving behavior Assessment System, designed to systematically evaluate drivers’ behavior and identify potential risks. Unlike conventional in-cabin driver monitoring systems, which generally rely on rule-based or supervised methods, TUN-DAS utilizes a unique time-series clustering approach. This method combines a
k
-means algorithm with Dynamic Time Warping (DTW) and DTW Barycenter Averaging (DBA), enabling a more nuanced and comprehensive assessment of driving behavior. The modification of conventional DTW in our system allows for effective computation of temporal differences between driving data. TUN-DAS stands out in its ability to detect both sporadic human errors and consistent unsafe driving patterns by analyzing anomalies within individual drivers and across a broader driver dataset. Our validation with bus driver data shows TUN-DAS’s effectiveness in driving behavior assessment, significantly achieving better accuracy than traditional rule-based evaluation methods and allowing for quantitative and temporal analysis of a series of driving behavior. This system holds promise for enhancing road safety by facilitating the early identification of high-risk drivers, thereby contributing to the development of safer driving environments.
Scoring driver behavior is crucial for profiling drivers to improve driving habits with beneficial feedback, significantly reducing the risks of accidents, fuel consumption, and traffic emissions. Although connected vehicles enable extensive collection of driving data for accurate scoring models, two key challenges persist: the absence of target scores for traditional regression algorithms and privacy concerns that hinder centralized learning to build unsupervised scoring models. We introduce FedDriveScore, a privacy-preserving unsupervised scoring framework tailored for the Internet of Vehicles context. This solution leverages a mixed distribution approach to form an objective scoring model that utilizes probability distributions and statistical weights to assess driver behavior. Homomorphic encryption-based federated learning is employed to collaboratively implement the scoring model across multiple vehicles without sharing raw data. To mitigate the detrimental effects of skewed metric distributions, an enhanced federated learning method is developed to ensure utility consistency between models generated through federated learning and centralized learning. Experimental results from two real-world datasets show that the proposed solution achieves remarkable performance, with scoring utility losses of less than 0.0001 (mean squared error) and 0.01 (mean absolute error), representing approximately a 99% improvement over traditional federated learning benchmark. Our solution demonstrates high alignment with centralized learning outcomes while effectively profiling drivers in an objective and equitable manner.
Intelligent Transportation Systems (ITS) aim to enhance traffic management by improving connectivity and data sharing among vehicles and road infrastructure. In a Mixed Connected and Connectionless Vehicles (MCCV) scenario consisting of connected vehicles equipped with On-Board Units (OBUs) and non-connected vehicles lacking OBUs, communication disparities create challenges in critical lane-changing overtaking decisions. These discrepancies hinder the adaptation of fully connected scenarios to dynamic interactions among these different types of vehicles. Considering the diversity in decision-making ways and capabilities of non-connected vehicles in MCCV scenarios, ensuring the coordinated execution of safe and efficient lane-changing overtaking maneuvers by multiple connected vehicles is crucial for enhancing traffic efficiency. Therefore, we propose a collaborative strategy to facilitate safer and more efficient lane-changing overtaking maneuvers for connected vehicles in the MCCV scenario. First, we design a multi-criteria priority detection, and a dynamic event-triggered mechanism based on confidence intervals to foster efficient collaboration among connected vehicles, optimizing decision-making and reducing conflicts. Second, to accommodate diverse driving styles of autonomous and human-driven vehicles, we introduce an Improved Dynamic Precise Fuzzy C-Means (IDP-FCM) algorithm to dynamically identify and adapt to different driving styles, thereby improving safety. Finally, tackling the challenge of multiple connected vehicles performing lane-changing overtaking involving hybrid action space, our proposed Multi-agent Contrastive Parameterized Dueling Deep Q-Network (MCPDDQN) algorithm incorporates contrastive learning to improve strategy stability in complex driving scenarios. Experimental results demonstrate the effectiveness of our strategy in improving road safety and traffic efficiency of the MCCV scenario.
Detecting abnormal driving behavior is critical for road traffic safety and the evaluation of drivers' behavior. With the advancement of machine learning (ML) algorithms and the accumulation of naturalistic driving data, many ML models have been adopted for abnormal driving behavior detection (also referred to in this paper as anomalies). Most existing ML-based detectors rely on (fully) supervised ML methods, which require substantial labeled data. However, ground truth labels are not always available in the real world, and labeling large amounts of data is tedious. Thus, there is a need to explore unsupervised or semi-supervised methods to make the anomaly detection process more feasible and efficient. To fill this research gap, this study analyzes large-scale real-world data revealing several abnormal driving behaviors (e.g., sudden acceleration, rapid lane-changing) and develops a Hierarchical Extreme Learning Machines (HELM) based semi-supervised ML method using partly labeled data to accurately detect the identified abnormal driving behaviors. Moreover, previous ML-based approaches predominantly utilized basic vehicle motion features (such as velocity and acceleration) to label and detect abnormal driving behaviors, while this study seeks to introduce Surrogate Measures of Safety (SMoS) as input features for ML models to improve the detection performance. Results from extensive experiments demonstrate the effectiveness of the proposed semi-supervised ML model with the introduced SMoS serving as important features. The proposed semi-supervised ML method outperforms other baseline semi-supervised or unsupervised methods regarding various metrics, e.g., delivering the best accuracy at 99.58% and the best F-1 measure at 0.9913. The ablation study further highlights the significance of SMoS for advancing the detection performance of abnormal driving behaviors.
To ensure the safety of Autonomous Vehicles (AVs), thorough testing across virtual simulation environments, closed facilities, and public roads is essential. Scenario-based testing stands out as a crucial method for evaluating AVs, with a key focus on constructing appropriate testing scenarios. Given the vulnerability of Powered Two-Wheelers (PTWs) riders, it is essential to investigate typical and representative car-to-PTWs crash scenarios and validate AV system safety performance in such situations. This study introduces a new method for generating high-risk scenarios by extracting typical testing scenarios from real-world crashes, thereby enhancing the realism of testing conditions. To evaluate the safety performance of AV systems, a crash-based testing approach is proposed. First, 222 car-to-PTWs crashes were extracted from the China In-depth Mobility Safety Study-Traffic Accident (CIMSS-TA) database and the pre-crash trajectories of crash-involved parties were accurately obtained by reconstructing the crashes case-by-case. Second, utilizing the
k
-medoids algorithm based on the Hausdorff distance to cluster these trajectories, six typical pre-crash trajectory clusters were extracted. Third, six sets of high-risk scenarios were generated using parameter discretization and combination testing based on the cluster centers. Finally, we conducted safety testing on the black-box automated driving system, Baidu Apollo, using the SVL Simulator virtual simulation platform. We evaluated its performance by subjecting it to the six sets of high-risk scenarios generated in this study. The experimental results demonstrate that Apollo can operate safely in most high-risk scenarios, indicating that automated driving systems can handle crashes that some human drivers cannot avoid, thereby improve traffic safety.
Transportation safety is highly correlated with driving behavior, especially human error playing a key role in a large portion of crashes. Modern instrumentation and computational resources allow for the monitorization of driver, vehicle, and roadway/environment to extract leading indicators of crashes from multi-dimensional data streams. To quantify variations that are beyond normal in driver behavior and vehicle kinematics, the concept of volatility is applied. The study measures driver-vehicle volatilities using the naturalistic driving data. By integrating and fusing multiple real-time streams of data, i.e., driver distraction, vehicular movements and kinematics, and instability in driving, this study aims to predict occurrence of safety critical events and generate appropriate feedback to drivers and surrounding vehicles. The naturalistic driving data is used which contains 7566 normal driving events, and 1315 severe events (i.e., crash and near-crash), vehicle kinematics, and driver behavior collected from more than 3500 drivers. In order to capture the local dependency and volatility in time-series data 1D-Convolutional Neural Network (1D-CNN), Long Short-Term Memory (LSTM), and 1DCNN-LSTM are applied. Vehicle kinematics, driving volatility, and impaired driving (in terms of distraction) are used as the input parameters. The results reveal that the 1DCNN-LSTM model provides the best performance, with 95.45% accuracy and prediction of 73.4% of crashes with a precision of 95.67%. Additional features are extracted with the CNN layers and temporal dependency between observations is addressed, which helps the network learn driving patterns and volatile behavior. The model can be used to monitor driving behavior in real-time and provide warnings and alerts to drivers in low-level automated vehicles, reducing their crash risk.
Connected and Automated Vehicles (CAVs) can potentially improve the performance of the transportation system by reducing human errors. This paper investigates the safety impact of CAVs in a mixed traffic with conventional vehicles at intersections. Analyzing real-world AV crashes in California revealed that rear-end crashes at intersections is the dominant crash type. Therefore, to enhance our understanding of the future interactions between human-driven vehicles with CAVs at intersections, a simulation framework was developed to model the mixed traffic environment of Automated Vehicles (AV), cooperative AVs, and conventional human-driven vehicles. In order to model AVs driving behavior, Adaptive Cruise Control (ACC) and cooperative ACC (CACC) models are utilized. Particularly, this study explores system improvements due to automation and connectivity across varying CAV market penetration scenarios. ACC and CACC car following models are used to mimic the behavior of AVs and cooperative AVs. Real-world connected vehicle data are utilized to modify and tune the acceleration/deceleration regimes of the Wiedemann model. Next, the driving volatility concept capturing variability in vehicle speeds was utilized to calibrate the simulation to represent the safety performance of a real-world environment. Two surrogate safety measures are used to evaluate the safety performance of a representative intersection under different market penetration rate of CAVs: the number of longitudinal conflicts and driving volatility. At low levels of ACC market penetration, the safety improvements were found to be marginal, but safety improved substantially with more than 40% ACC penetration. Additional safety improvements can be achieved more quickly through the addition of cooperation and connectivity through CACC. Furthermore, ACC/CACC vehicles were found to improve mobility performance in terms of average speed and travel time at intersections.
Abstract: This report focuses on safety aspects of connected and automated vehicles (CAVs). The fundamental question to be answered is how can CAVs improve road users' safety? Using advanced data mining and thematic text analytics tools, the goal is to systematically synthesize studies related to Big Data for safety monitoring and improvement. Within this domain, the report systematically compares Big Data initiatives related to transportation initiatives nationally and internationally and provides insights regarding the evolution of Big Data science applications related to CAVs and new challenges. The objectives addressed are:
• Creating a database of Big Data efforts by acquiring reports, white papers, and journal publications;
• Applying text analytics tools to extract key concepts, and spot patterns and trends in Big Data initiatives;
• Understanding the evolution of CAV Big Data in the context of safety by quantifying granular taxonomies and modeling entity relations among contents in CAV Big Data research initiatives, and
• Developing a foundation for exploring new approaches to tracking and analyzing CAV Big Data and related innovations.
The study synthesizes and derives high-quality information from innovative research activities undertaken by various research entities through Big Data initiatives. The results can provide a conceptual foundation for developing new approaches for guiding and tracking the safety implications of Big Data and related innovations.
Obtaining accurate speed and travel time information is a challenge for researchers, geographers, and transportation agencies. In the past, traffic data were usually acquired and disseminated by government agencies through fixed-location sensors. High costs, infrastructure demands, and low coverage levels of these sensor devices require agencies and researchers to look beyond the traditional approaches. With the emergence of smartphones and navigation apps, location-based and crowdsourced Big Data are receiving increased attention. In this regard, location-based big data (LocBigData) collected from probe vehicles and road users can be used to provide speed and travel time information in different locations. Examining the quality of crowdsourced data is essential for researchers and agencies before using them. This study assessed the quality of Waze speed data from surface streets and conducted a case study in Sevierville, Tennessee. Typically, examining the quality of these data in surface streets and arterials is more challenging than freeways data. This research used Bluetooth speed data as the ground truth, which is independent of Waze data. In this study, three steps of methodology were used. In the first step, Waze speed data was compared to Bluetooth data in terms of accuracy, mean difference, and distribution similarity. In the second step, a k-means algorithm was used to categorize Waze data quality, and a multinomial logistics regression model was performed to explore the significant factors that impact data quality. Finally, in the third step, machine learning techniques were conducted to predict the data quality in different conditions. The result of the comparison showed a similar pattern and a slight difference between datasets, which verified the quality of Waze speed data. The statistical model indicates that that Waze speed data are more accurate in peak hours than in night hours. Also, the traffic speed, traffic volume, and segment length have a significant association on the accuracy of Waze data on surface streets. Finally, the result of machine learning prediction showed that a KNN method performed the highest prediction accuracy of 84.5% and 82.9% of the time for training and test datasets, respectively. Overall, the study results suggest that Waze speed data is a promising data source for surface streets.
While the cost of crashes exceeds $1 Trillion a year in the U.S. alone, the availability of high-resolution naturalistic driving data provides an opportunity for researchers to conduct an in-depth analysis of crash contributing factors, and design appropriate interventions. Although police-reported crash data provides information on crashes, this study takes advantage of the SHRP2 Naturalistic Driving Study (NDS) which is a unique dataset that allows new insights due to detailed information on driver behavior in normal, pre-crash, and near-crash situations, in addition to trip and vehicle performance characteristics. This paper investigates the role of pre-crash driving instability, or driving volatility, in crash intensity (measured on a 4-point scale from a tire-strike to an injury crash) by analyzing microscopic vehicle kinematic data. NDS data are used to investigate not only the vehicle movements in space but also the instability of vehicles prior to the crash and their contribution to crash intensity using path analysis. A subset of the data containing 617 crash events with around 0.18 million temporal trajectories are analyzed. To quantify driving instability, microscopic variations or volatility in vehicular movements before a crash are analyzed. Specifically, nine measures of pre-crash driving volatility are calculated and used to explain crash intensity. While most of the measures are significantly correlated with crash intensity, substantial positive correlations are observed for two measures representing speed and deceleration volatilities. Modeling results of the fixed and random parameter probit models revealed that volatility is one of the leading factors increasing the probability of a severe crash. Additionally, the speed prior to a crash is highly correlated with intensity outcomes, as expected. Interestingly, distracted and aggressive driving are highly correlated with driving volatility and have substantial indirect effects on crash intensity. With volatile driving serving as a leading indicator of crash intensity, given the crashes analyzed in this study, early warnings and alerts for the subject vehicle driver and proximate vehicles can be helpful when volatile behavior is observed.
Connected and automated vehicles have enabled researchers to use big data for development of new metrics that can enhance transportation safety. Emergence of such a big data coupled with computational power of modern computers have enabled us to obtain deeper understanding of instantaneous driving behavior by applying the concept of “driving volatility” to quantify variations in driving behavior. This paper brings in a methodology to quantify variations in vehicular movements utilizing longitudinal and lateral volatilities and proactively studies the impact of instantaneous driving behavior on type of crashes at intersections. More than 125 million Basic Safety Message data transmitted between more than 2800 connected vehicles were analyzed and integrated with historical crash and road inventory data at 167 intersections in Ann Arbor, Michigan, USA. Given that driving volatility represents the vehicular movement and control, it is expected that erratic longitudinal/lateral movements increase the risk of crash. In order to capture variations in vehicle control and movement, we quantified and used 30 measures of driving volatility by using speed, longitudinal and lateral acceleration, and yaw-rate. Rigorous statistical models including fixed parameter, random parameter, and geographically weighted Poisson regressions were developed. The results revealed that controlling for intersection geometry and traffic exposure, and accounting unobserved factors, variations in longitudinal control of the vehicle (longitudinal volatility) are highly correlated with the frequency of rear-end crashes. Intersections with high variations in longitudinal movement are prone to have higher rear-end crash rate. Referring to sideswipe and angle crashes, along with speed and longitudinal volatility, lateral volatility is substantially correlated with the frequency of crashes. When it comes to head-on crashes, speed, longitudinal and lateral acceleration volatilities are highly associated with the frequency of crashes. Intersections with high lateral volatility have higher risk of head-on collisions due to the risk of deviation from the centerline leading to head-on crash. The developed methodology and volatility measures can be used to proactively identify hotspot intersections where the frequency of crashes is low, but the longitudinal/lateral driving volatility is high. The reason that drivers exhibit higher levels of driving volatility when passing these intersections can be analyzed to come up with potential countermeasures that could reduce volatility and, consequently, crash risk.
Heterogeneity of crash data masks the underlying crash patterns and perplexes crash analysis. This paper aims to explore an
advanced high-dimensional clustering approach to investigate heterogeneity in large datasets. Detailed records of crashes
involving large trucks occurring in the state of Florida between 2007 and 2016 were examined to identify truck crash patterns
and significant conditions contributing to the patterns. The block clustering method was applied to more than 220,000 crash
records with nearly 200 attributes. The analysis showed promising results in segmenting a large heterogeneous dataset into
meaningful subgroups (with 95.72% average degree of homogeneity for selected blocks). The goodness of fit for clustering
methods is evaluated and both integrated completed likelihood (ICL) and pseudo-likelihood values improved significantly
(20.8% and 21.1% respectively). Attribute clustering showed distinct characteristics for each cluster. Crash clustering revealed
significant differences among the clusters and suggested that this crash dataset could be portioned as same-direction, opposing-direction, and single-vehicle crashes. Individual blocks defined by both row and column clustering were further investigated to better understand the contribution set of conditions that lead to large truck crashes. Major features for each of the
three major types of crashes were analyzed, which may provide additional insights to develop potential countermeasures and
strategies that target specific segments. The clustering approach could be used as a preanalysis method to identify homogeneous subgroups for further analysis, which will help enhance the effectiveness of safety programs.
With the emergence of high-frequency connected and automated vehicle data, analysts have become able to extract useful information from them. To this end, the concept of "driving volatility" is defined and explored as deviation from the norm. Several measures of dispersion and variation can be computed in different ways using vehicles' instantaneous speed, acceleration, and jerk observed at intersections. This study explores different measures of volatility, representing newly available surrogate measures of safety, by combining data from the Michigan Safety Pilot Deployment of connected vehicles with crash and inventory data at several intersections. The intersection data was error-checked and verified for accuracy. Then, for each intersection, 37 different measures of volatility were calculated. These volatilities were then used to explain crash frequencies at intersection by estimating fixed and random parameter Poisson regression models. Results show that an increase in three measures of driving volatility are positively associated with higher intersection crash frequency, controlling for exposure variables and geometric features. More intersection crashes were associated with higher percentages of vehicle data points (speed & acceleration) lying beyond threshold-bands. These bands were created using mean plus two standard deviations. Furthermore, a higher magnitude of time-varying stochastic volatility of vehicle speeds when they pass through the intersection is associated with higher crash frequencies. These measures can be used to locate intersections with high driving volatilities, i.e., hot-spots where crashes are waiting to happen. Therefore, a deeper analysis of these intersections can be undertaken and proactive safety countermeasures considered at high volatility locations to enhance safety.
Since the advantage of hidden Markov model in dealing with time series data and for the sake of identifying driving style, three driving style (aggressive, moderate and mild) are modeled reasonably through hidden Markov model based on driver braking characteristics to achieve efficient driving style. Firstly, braking impulse and the maximum braking unit area of vacuum booster within a certain time are collected from braking operation, and then general braking and emergency braking characteristics are extracted to code the braking characteristics. Secondly, the braking behavior observation sequence is used to describe the initial parameters of hidden Markov model, and the generation of the hidden Markov model for differentiating and an observation sequence which is trained and judged by the driving style is introduced. Thirdly, the maximum likelihood logarithm could be implied from the observable parameters. The recognition accuracy of algorithm is verified through experiments and two common pattern recognition algorithms. The results showed that the driving style discrimination based on hidden Markov model algorithm could realize effective discriminant of driving style.
Supervised learning approaches are widely used for driving style classification; however, they often require a large amount of labeled training data, which is usually scarce in a real-world setting. Moreover, it is time-consuming to manually label huge amounts of driving data due to uncertainties of driver behavior and variances among the data analysts. To address this problem, a semi-supervised approach, a semi-supervised support vector machine (S3VM), is employed to classify drivers into aggressive and normal styles based on a few labeled data points. First, a few data clusters are selected and manually labeled using a k-means clustering method. Then, a specific differentiable surrogate of a loss function is developed, which makes it feasible to use standard optimization tools to solve the non-convex optimization problem. One of the most popular quasi-Newton algorithms is then used to assign the optimal label to all of the training data. Lastly, we compare the S3VM method with a support vector machine (SVM) method for classifying driving styles from different amounts of labeled data. Experiments show that the S3VM method can improve the classification accuracy by about 10% and reduce the labeling effort by using only a few labeled data clusters amongst huge amounts of unlabeled data.
Big Data has emerged with new opportunities for research, development, innovation, and business. It is characterized by the so-called four Vs: volume, velocity, veracity, and variety, and it may bring significant value through the processing of a large amount of data. The transformation of Big Data's four Vs into the fifth V (value) is a grand challenge for processing capacity. Cloud computing has emerged as a new paradigm to provide computing as a utility service for addressing different processing needs with (a) on-demand services, (b) pooled resources, (c) elasticity, (d) broadband access, and (e) measured services. The utility of delivering computing capability fosters a potential solution for the transformation of Big Data's four Vs into the fifth V. This paper investigates how cloud computing can be utilized to address Big Data challenges to enable such transformation. We introduce and review four geospatial scientific examples, including climate studies, geospatial knowledge mining, land-cover simulation, and dust storm modeling. The method is presented in a tabular framework as a guidance to leverage cloud computing for Big Data solutions. It is illustrated that the framework method supports the life cycle of Big Data processing, including management, access, mining analytics, simulation, and forecasting. This tabular framework can also be referred as a guidance to develop potential solutions for other big geospatial data challenges and initiatives such as smart cities.
The aim of this study was to validate the stability of the different factors of the Multidimensional Driving Style Inventory (MDSI) [16], which was originally developed and validated with participants in different geographical areas of Israel. In this study, the questionnaire was distributed in the Netherlands and Belgium. A factor analysis of the data of 364 participants revealed five of the eight factors that resulted from the original factor analysis: Angry driving, Anxious driving, Dissociative driving, Distress-reduction driving, and Careful driving style. In addition, 24 items divided over the five factors seem to be stable compared to the 44 items divided over the eight factors of the original analysis. The factors revealed through the analysis of these data were used to determine driver profiles, consisting of one or two driving styles. The next step is to compare self-report data on driving style to actual driving behaviour.
In this paper the various driving style analysis solutions are investigated. An in-depth investigation is performed to identify the relevant machine learning and artificial intelligence algorithms utilised in current driver behaviour and driving style analysis systems. This review therefore serves as a trove of information, and will inform the specialist and the student regarding the current state of the art in driver style analysis systems, the application of these systems and the underlying artificial intelligence algorithms applied to these applications. The aim of the investigation is to evaluate the possibilities for unique driver identification utilizing the approaches identified in other driver behaviour studies. It was found that Fuzzy Logic inference systems, Hidden Markov Models and Support Vector Machines consist of promising capabilities to address unique driver identification algorithms if model complexity can be reduced.
Today's smartphones and mobile devices typically embed advanced motion sensors. Due to their increasing market penetration, there is a potential for the development of distributed sensing platforms. In particular, over the last few years there has been an increasing interest in monitoring vehicles and driving data, aiming to identify risky driving maneuvers and to improve driver efficiency. Such a driver profiling system can be useful in fleet management, insurance premium adjustment, fuel consumption optimization or CO2 emission reduction. In this paper, we analyze how smartphone sensors can be used to identify driving maneuvers and propose SenseFleet, a driver profile platform that is able to detect risky driving events independently from the mobile device and vehicle. A fuzzy system is used to compute a score for the different drivers using real-time context information like route topology or weather conditions. To validate our platform, we present an evaluation study considering multiple drivers along a predefined path. The results show that our platform is able to accurately detect risky driving events and provide a representative score for each individual driver.
Detection and classification of aggressive driving can be based on the use of physiological signals or biometric information like electrocardiogram, electro dermal activity, and respiration. This research proposes a driving performance inference system based on the signature of acceleration in the two dimensions and speed. Driving style can be categorized to: below normal, normal, aggressive, and very aggressive. One of the targets of this paper is to recognize the driving events that fall into each of these categories. Many of driving styles are a major cause of traffic crashes. Many of these styles can be detected with reasonable accuracy using driving performance. The main idea is to utilize the 2-axis accelerometer that is embedded in most of the GPS tracker devices that are used in vehicle tracking and fleet management to recognize the driving styles. This method can be utilized for vehicle active safety purpose. Fuzzy logic inference system is used in classification of the extracted feature to the predefined driving styles. the power of the Euclidean norm of longitudinal and lateral is used as input to the fuzzy inference system in addition to the speed. Classification of the driving style is real time and almost cost nothing.
The rapid development and increasing availability of various location acquisition technologies provide geospatial studies with both opportunities and challenges. These opportunities and challenges are discussed in this paper focusing on the following three aspects: the massive acquisition of location data and data quality, the analysis of massive location data and pattern discovery, and privacy protection for massive location data. This paper examines the current status of and the potential opportunities for geospatial research in these three areas and notes the major challenges. Finally, the development of this special issue is described, and the four articles included in this special issue are presented.
This paper illustrates the use of neural network techniques for analyzing headway data collected from a group of 36 driving subjects during normal on-highway driving. Pattern recognition methods are used to identify different types of headway-keeping behavior exhibited by these drivers and their relative distributions. Possibilities for using neural networks to represent longitudinal control behavior of drivers are also considered and discussed.
Research for two-lane rural road driver passing maneuver evaluation using STSIM, an interactive driving simulator, collected data is presented by the authors. Self-reported questionnaires were used to collect drivers' socioeconomic characteristics and driving style indicators, and simulator experiments were used to obtain driving behavior observations. Driving a 9.5 km rural road section with two-lanes and no intersections was requested of participants. Simulated vehicles and the subject vehicles' positions and speeds were recorded at a 0.1 second resolution. Experiment-collected data was used for driver passing decision explanation model development. Study results indicate that subject vehicle speed and relation to vehicle it is passing are the most important passing behavior factors. Passing decisions are also affected by drivers' driving styles and socio-demographic characteristics.
This paper describes the influence on vehicle emissions and energy consumption of different vehicle parameters and driving style as well as of traffic measures taken in order to increase transport safety or to reduce traffic jams. This should allow the Flemish Regional Government to perform more realistic modelling of the impact of transport on air pollution.
The methodology is based on on-road measurements, roll-bench emission tests, vehicle simulations and regional emission modelling (for the Flemish Region, which encompasses the northern part of Belgium and is one of three entities that constitute the Federal Kingdom of Belgium). A vehicle simulation programme (VSP) has assisted in the assessment of the individual vehicle parameters (weight, gear shifts, tyre pressure, etc.). Different drive styles (sportive, EcoDriving, etc.) were measured on-road and evaluated on a roll-bench. Typical speed profiles corresponding to different traffic measures such as roundabouts, phased traffic lights, etc., were also recorded at different locations in the Flemish Region. All data were distilled into small driving cycles, representative of a certain traffic situation or driving style, and repeated on a roll-bench to measure the emissions in controlled circumstances.
Technical solutions as well as educational programmes are proposed as possible measures to reduce the influence of driving style on emissions and fuel consumption. Finally, the results indicate how the emission calculations performed by the Flemish government can be improved.
There are many systems to evaluate driving style based on smartphone sensors without enough awareness from the context. To cover this gap, we propose a new system namely CADSE system to consider the effects of traffic levels and car types on driving evaluation. CADSE system includes three subsystems to calibrate smartphone, to classify the maneuvers, and to evaluate driving styles. For each maneuver, the smartphone sensors data are gathered in three successive time intervals referred as pre-maneuver, in-maneuver, and post-maneuver times. Then, we extract some important mathematical and experimental features from these data. Afterwards, we propose an ensemble learning method on these features to classify the maneuvers. This ensemble method includes decision tree, support vector machine, multi-layer perceptron, and k-nearest neighbors. Finally, we develop a rule-based fuzzy inference system to integrate the outputs of these algorithms and to recognize dangerous and safe maneuvers. CADSE saves this result in driver’s profile to consider more for dangerous driving recognition. The experimental results show that accuracy, precision, recall, and F-measure of CADSE system are greater than 94%, 92%, 92%, and 93%, respectively that prove the system efficiency.
Connected and automated vehicle technologies have the potential to significantly improve transportation system performance. In particular, advanced driver-assistance systems, such as adaptive cruise control (ACC) and cooperative adaptive cruise control (CACC), may lead to substantial improvements in performance by decreasing driver inputs and taking over control of the vehicle. However, the impacts of these technologies on the vehicle- and system-level energy consumption, emissions, and safety have not been quantified in field tests. The goal of this paper is to study the impacts of automated and cooperative systems in mixed traffic containing conventional, ACC, and CACC vehicles. To reach this goal, experimental data based on real-world conditions are collected (in tests conducted by the Federal Highway Administration and the U.S. Department of Transportation) with presence of ACC, CACC, and conventional vehicles in a vehicle platoon scenario and a cooperative merging scenario. Specifically, a platoon of five vehicles with different vehicle type combinations is analyzed to generate new knowledge about potential safety, energy efficiency, and emission improvement from vehicle automation and cooperation. Results show that adopting the CACC system in a five-vehicle platoon substantially reduces the driving volatility and reduces the risk of rear-end collision which consequently improves safety. Furthermore, it decreases fuel consumption and emissions compared with the ACC system and manually-driven vehicles. Results of the merging scenario show that while the cooperative merging system slightly reduces the driving volatility, the fuel consumption and emissions can increase because of sharper accelerations of CACC vehicles compared with manually-driven vehicles.
This study contrasted the performance of drivers under actual and simulated driving conditions, in order to assess the validity of the simulators and test the hypothesis that driving is composed of largely orthogonal sub-tasks. Thirty experienced drivers completed an on-road driving test and drove two different simulators, each simulator drive comprising seven difficulty-moderated driving scenarios. Between-simulator contrasts revealed largely absolute validity, the anticipated effects of increased difficulty within driving scenarios, but weak relationships between performance of different driving scenarios. On-road driving was reliably assessed by a nationally-recognised expert driving assessor, as reflected by standard statistical measures of reliability and consistency. However, on-road driving revealed relatively little cross-category correlation of on-road driving errors, or between on-road and simulator driving. Thus, despite the compelling evidence of absolute and relative validity within and between simulators, there is little evidence of criterion validity (i.e. relationship to on road driving, as assessed by the expert assessor). Moreover, the study provides strong evidence for orthogonality in the driving task- driving comprises large numbers of relatively separate tasks.
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.
The current study aims to present a model to characterize changes in network traffic flows as a result of implementing connected and autonomous vehicle (CAV) technology based on traffic network and built-environment characteristics. To develop such a model, first, POLARIS agent-based modeling platform is used to predict changes in average daily traffic (ADT) under CAVs scenario in the road network of Chicago metropolitan area as the dependent variable of the model. Second, a comprehensive set of variables and indicators representing network characteristics and urban structure patterns are generated. Three machine learning models namely K-Nearest neighbors, Random Forest, and eXtreme Gradient Boosting are developed and validated to establish the relationship between network characteristics and changes in ADT under CAVs scenario. The estimated models are found to yield acceptable performance. In addition, SHapley Additive exPlanations (SHAP) analysis tool is employed to investigate the impact of important features on changes in ADT, which discloses the most important link properties, network features, and demographic information in predicting change in ADT under the analyzed CAVs scenario.
The Multidimensional Driving Style Inventory or MDSI constitutes, perhaps, the most relevant
tool for measuring driving styles. Since its releasing in 2004, it has been applied
worldwide to different samples of drivers, showing an important value and utility for road
safety. However, empirical studies using the MDSI on professional drivers are scarce and, to
the date, there is no validated version of the instrument in this workforce yet. Objectives:
This study had two aims. First, to describe in detail the validation of the Taubman-Ben-Ari’s
MDSI among professional drivers and, second, to test its convergent validity with other key
relevant factors present in the work environment of this particular population: driving
anger, job strain and occupational driving crashes. Method: The data used for this validation
was gathered from a representative sample of 752 Colombian professional drivers and
analyzed by means of competitive Confirmatory Factor Analyses (CFAs), assessing psychometric
properties and obtaining an optimized structure for the instrument applied to active
transportation workers. Results: The outcomes of this study suggest a clear factorial structure,
adequate model fit, factorial weights, reliability and internal consistency, keeping the
re-evaluated four-factor structure of the questionnaire: Reckless & Careless (F1); Anxious
(F2); Angry & Hostile (F3); and Patient & Careful (F4). Conclusion: This applied research
supports the hypothesis that the validated version of MDSI in professional drivers, together
with further measures applied to other work environment factors, may play a relevant role
in the improvement of driving safety and injury prevention for this vulnerable workforce
from the perspective of occupational research in transportation.
The state-of-the-practice for municipal traffic agencies identifying high-risk road segments has been to use data on prior crashes. While historical traffic crash data is valuable in improving roadway safety, it relies on prior observations rather than future crash likelihoods. Recently, however, researchers have developed predictive crash methods based on "abnormal driving events." These include abrupt and atypical vehicle movements indicative of crash avoidance maneuvers and/or near-crashes, especially on highways. Due to limited data, the previous research only tested the crash-jerk ratio function on highways but not on other types of roads. This paper describes research that used naturalistic driving data collected from global positioning system (GPS) sensors to locate high concentrations of abrupt and atypical vehicle movements based on vehicle acceleration and vehicle rate of change of acceleration (jerk) on two interrupted highways. Statistical analyses revealed that clusters of high magnitude jerk events while decelerating were significantly correlated to long-term crash rates at these locations. These significant and consistent relationships between jerks and crashes suggest that such observational data can be used as surrogate measures of safety and as a way of predicting safety problems, further improving crash prediction models.
Based on on-board diagnostics and Global Position System installed in taxicabs, driver behavior data is collected. Left turn data on six similar curves are extracted, and speed, acceleration, yaw rate, and sideslip angle of drivers in time series are selected as clustering indexes. Initial clustering is implemented by Dynamic Time Warping (DTW) and Hierarchical Clustering, and the clustering results are put into the Hidden Markov Model (HMM) to iteratively optimize the results for achieving convergence. Driver behavior patterns over time while driving on the curves and the statistical characteristics of different groups are examined. All indexes including lateral vehicle control and longitudinal vehicle control have a significant difference in different groups, indicating that the clustering method of DTW and HMM can effectively classify driver behavior. Finally, the driving behavior in different groups is further investigated and classified based on characteristics related to safe and ecological driving. This method can be applied by automobile insurance companies, and for the development of specific training courses for drivers to optimize their driving behavior.
The paper focuses on a task of stochastic modeling the driving style and its online estimation while driving. The driving style is modeled by means of a mixture model with normal and categorical components as well as a data-dependent pointer. The mixture parameters and the actual driving style are estimated with the help of a recursive algorithm under the Bayesian methodology. The main contributions of the presented approach are: (i) the online estimation of the driving style while driving, taking into account data up to the current time instant; (ii) the joint model for continuous and discrete data measured on a vehicle; (iii) the data-dependent model of the driving style conditioned by the values of fuel consumption; (iv) the use of the model both for detection of clusters according to the driving style and prediction of the fuel consumption along with other variables; and (v) the universal modeling with the help of mixtures, which allows us to use different combinations of components and pointer models as well as to specify the initialization approach suitable for the considered problem. Results of the driving style detection in real measurements and comparison with the theoretical counterparts are demonstrated.
Knowledge about the driving behavior of a driver is important for applications in many different areas, especially for Advanced Driver Assistance Systems. The driving style does not only affect the current driver and his vehicle but also his environment. For example, usage-based insurances classify the driving style in order to reward calm drivers by granting them a discount. In this paper we present a novel algorithm to provide an accurate classification of a person's driving style. Our model is based on the identification of driving maneuvers and the classification of the driving style for these events using artificial neural networks. Furthermore, an overall score of the driving style for one trip is calculated based on the classified events. We validate our developed model in 58 test trips from different test drivers using a recently developed low-cost measuring device based on a Raspberry Pi.
The results of our validation show that the model can identify more than 90 % of the driving maneuvers correctly. Moreover, the driving style classification matches the assessment of the driver in 81 % of the relevant trips with a normalized average mean squared error of less than 11 %. In addition, a moving average of the calculated score for each event shows validated changes in the driving behavior of the test persons.
Urban road traffic is highly dynamic. Traffic conditions vary in time and with location and so do the movement patterns of individual road users. In this article, a movement pattern is the behaviour of a car when traversing a road link in an urban road network. A movement pattern can be recorded with a global navigation satellite system (GNSS), such as the Global Positioning System (GPS). A movement pattern has a specific energy-efficiency, which is a measure of how fuel-intensively the car is moving. For example, a car driving uniformly at medium speed consumes little fuel and, therefore, is energy-efficient, whereas stop-and-go driving consumes much fuel and is energy-inefficient. In this article we introduce a model to estimate the energy-efficiency of movement patterns in urban road traffic from GNSS data. First, we derived statistical features about the car's movement along the road. Then, we compared these to fuel consumption data from the car's controller area network (CAN) bus, normalized to the car's overall range of fuel consumption. We identified the optimal feature set for prediction. With the optimal feature set we trained, tested and verified a model to estimate energy-efficiency, with the fuel consumption serving as ground truth. Existing fuel consumption models usually view movement as a snapshot. Thus, the behaviour of the car remains unknown that causes a movement pattern to be energy-efficient or energy-inefficient. Our model views movement as a process and allows to interpret this process. A movement pattern can, for example, be energy-inefficient because the car is driving in stop-and-go traffic, because it is travelling at high speed, or because it is accelerating. Our model allows to distinguish between these different types of behaviours. Thus, it can provide new insights into the dynamics of urban road traffic and its energy-efficiency.
When vehicles share their status information with other vehicles or the infrastructure, driving actions can be planned better, hazards can be identified sooner, and safer responses to hazards are possible. The Safety Pilot Model Deployment (SPMD) is underway in Ann Arbor, Michigan; the purpose is to demonstrate connected technologies in a real-world environment. The core data transmitted through Vehicle-to-Vehicle and Vehicle-to Infrastructure (or V2V and V2I) applications are called Basic Safety Messages (BSMs), which are transmitted typically at a frequency of 10 Hz. BSMs describe a vehicle's position (latitude, longitude, and elevation) and motion (heading, speed, and acceleration). This study proposes a data analytic methodology to extract critical information from raw BSM data available from SPMD. A total of 968,522 records of basic safety messages, gathered from 155 trips made by 49 vehicles, was analyzed. The information extracted from BSM data captured extreme driving events such as hard accelerations and braking. This information can be provided to drivers, giving them instantaneous feedback about dangers in surrounding roadway environments; it can also provide control assistance. While extracting critical information from BSMs, this study offers a fundamental understanding of instantaneous driving decisions. Longitudinal and lateral accelerations included in BSMs were specifically investigated. Varying distributions of instantaneous longitudinal and lateral accelerations are quantified. Based on the distributions, the study created a framework for generating alerts/warnings, and control assistance from extreme events, transmittable through V2V and V2I applications. Models were estimated to untangle the correlates of extreme events. The implications of the findings and applications to connected vehicles are discussed in this paper.
When vehicles share their status information with other vehicles or the infrastructure, driving actions can be planned better, hazards can be identified sooner, and safer responses to hazards are possible. The Safety Pilot Model Deployment (SPMD) is underway in Ann Arbor, Michigan; the purpose is to demonstrate connected technologies in a real-world environment. The core data transmitted through Vehicle-to-Vehicle and Vehicle-to-Infrastructure (or V2V and V2I) applications are called Basic Safety Messages (BSMs), which are transmitted typically at a frequency of 10 Hz. BSMs describe a vehicle’s position (latitude, longitude, and elevation) and motion (heading, speed, and acceleration). This study proposes a data analytic methodology to extract critical information from raw BSM data available from SPMD. A total of 968,522 records of basic safety messages, gathered from 155 trips made by 49 vehicles, was analyzed. The information extracted from BSM data captured extreme driving events such as hard accelerations and braking. This information can be provided to drivers, giving them instantaneous feedback about dangers in surrounding roadway environments; it can also provide control assistance. While extracting critical information from BSMs, this study offers a fundamental understanding of instantaneous driving decisions. Longitudinal and lateral accelerations included in BSMs were specifically investigated. Varying distributions of instantaneous longitudinal and lateral accelerations are quantified. Based on the distributions, the study created a framework for generating alerts/warnings, and control assistance from extreme events, transmittable through V2V and V2I applications. Models were estimated to untangle the correlates of extreme events. The implications of the findings and applications to connected vehicles are discussed in this paper.
*Note*: Abstract will be updated to be consistent with the final version.
Nowadays more and more driver assistance systems are implemented in cars. By adapting the system to the driving style of the driver, the acceptance of the driver to such a system could be enhanced. In this paper a system for online driving style recognition is designed. It is implemented in Matlab/Simulink and uses fuzzy logic for identifying the current driving style. It is fully parameterisable via a central parameter file and could therefore be adapted to nearly every car. The recognition was tested by using a vehicle dynamics simulation with 68% correct classifications over time.
Given the increase of vehicles in traffic, the traffic accidents became a crucial and urgent issue for the countries. Particularly, in heavy traffic conditions, rear-end collisions are the majority of traffic accidents, which make the traffic jam worse. This paper proposes a novel approach to rear-end collision warning systems using areas of license plates acquired with a single camera mounted on a car. The edges of the front car's license plate are segmented and a rectangle is sketched to calculate the area which is used for estimating distance between the cars. Relative speed of the front car is computed using the differences of the rectangles in a specific time. Distance and relative speed are obtained from the estimated areas of the license plates and transferred to the fuzzy inference system to send a warning signal to the driver for collision prevention, in emergency cases. The experiments are greatly encouraging that the number plate segmentation can be utilized to estimate the distance and fuzzy inference system can be developed to create a warning signal to the drivers.
Accurately understanding driving behaviour is of crucial importance for advanced driving assistant systems such as adaptive cruise control system and intelligent forward collision warning system. To understand different driving styles, this study employs the clustering method and topic model to extract latent driving states, which can elaborate and analyse the commonness and individuality of driving behaviour characteristics with the longitudinal driving behaviour data collected by the instrumented vehicle. To handle the large set of data and discover the valuable knowledge, the data mining techniques including ensemble clustering method based on the kernel fuzzy C-means algorithm and the modified latent Dirichlet allocation model are employed in this study. The 'aggressive', 'cautious' and 'moderate' driving states are discovered and the underlying quantified structure is built for the driving style analysis.
Driving styles can be broadly characterized as calm or volatile, with significant implications for traffic safety, energy consumption and emissions. How to quantify the extent of calm or volatile driving and explore its correlates is a key research question investigated in the study. This study contributes by leveraging a large-scale behavioral database to analyze short-term driving decisions and develop a new driver volatility index to measure the extent of variations in driving. The index captures variation in driving behavior constrained by the performance of the vehicle from a decision-making perspective. Specifically, instantaneous driving decisions include maintaining speed, accelerating, decelerating, maintaining acceleration/deceleration, or jerks to vehicle, i.e., the decision to change marginal rate of acceleration or deceleration. A fundamental understanding of instantaneous driving behavior is developed by categorizing vehicular jerk reversals (acceleration followed by deceleration), jerk enhancements (increasing accelerations or decelerations), and jerk mitigations (decreasing accelerations or decelerations). Volatility in driving decisions, captured by jerky movements, is quantified using data collected in Atlanta, GA during 2011. The database contains 51,370 trips and their associated second-by-second speed data, totaling 36 million seconds. Rigorous statistical models explore correlates of volatility that include socioeconomic variables, travel context variables, and vehicle types. The study contributes by proposing a framework that is based on defining instantaneous driving decisions in a quantifiable way using big data generated by in-vehicle GPS devices and behavioral surveys.
This paper presents an innovative idea for the classification of individual drivers. The classification is based on each driver's driving features like, ratio of indicators to turns, number of brakes, number of time horn used, average gear, average speed, maximum speed and gear. K-means and hierarchical clustering is used to separate out the slow, normal and fast driving styles based on recorded data. Experimental result shows that k-means outperformed hierarchical clustering for recorded multi-attribute data.
This research effort aims to investigate the hypothesis that drivers apply different driving styles in their daily driving tasks. A two-step algorithm is used for segmentation and clustering. First, a car-following period is broken into different duration segments that account for their temporal distribution. Second, the segments produced by the previous step are clustered based on similarity. Variations of k-means clustering and optimization techniques are used in this process. The segments centroids, used for clustering, are 8-dimensional and are produced by taking the average of the data points in each segment based on longitudinal acceleration, lateral acceleration, gyro (yaw rate), vehicle speed, lane offset, gamma (yaw angle), range, and range rate. The results of this methodology are continuous segments of car-following behavior as well as clusters of segments that show similar data and thus similar behaviors. The sample used in this paper included three different truck drivers that are representative of a high-risk driver, a medium-risk driver, and a low-risk driver. . In summary, the results revealed behavior that changed within a car-following period, between car-following periods, and between drivers. Each driver showed a unique distribution of behavior, but some of the behaviors existed in more than one driver but at different frequencies.
In recent year many researcher and industries are working on VANET and trying to implement the concepts in real world. Many VANET systems are proposed and tested on simulation but very few inventors implemented it. Vehicular ad hoc networks (VANETs) are being developed to provide on-demand wireless communication infrastructure among vehicles and authorities. Such an infrastructure is expected to deliver multiple road safety and driving assistance applications. Vehicles will be equipped with sensors and communication devices that will allow them to cooperate with each other. Vehicles can exchange different type of information as per requirements on demand for specified application. With the purpose of supporting and improving data collection and distribution, in this paper a smart phone-based platform is designed that exploits low-cost dedicated hardware to interact with sensors on board and in the vehicle surroundings.
Driving aggressively increases the risk of accidents. Assessing a person's driving style is a useful way to guide aggressive drivers toward having safer driving behaviors. A number of studies have investigated driving style, but they often rely on the use of self-reports or simulators, which are not suitable for the real-time, continuous, automated assessment and feedback on the road. In order to understand and model aggressive driving style, we construct an in-vehicle sensing platform that uses a smartphone instead of using heavyweight, expensive systems. Utilizing additional cheap sensors, our sensing platform can collect useful information about vehicle movement, maneuvering and steering wheel movement. We use this data and apply machine learning to build a driver model that evaluates drivers' driving styles based on a number of driving-related features. From a naturalistic data collection from 22 drivers for 3 weeks, we analyzed the characteristics of drivers who have an aggressive driving style. Our model classified those drivers with an accuracy of 90.5% (violation-class) and 81% (questionnaire-class). We describe how, in future work, our model can be used to provide real-time feedback to drivers using only their current smartphone.
In this work the trade-off between economic, therefore fuel saving, and ecologic, pollutant emission reducing, driving is discussed. The term eco-driving is often used to refer to a vehicle operation that minimizes energy consumption. However, for eco-driving to be environmentally friendly not only fuel consumption but also pollutant emissions should be considered. In contrast to previous studies, this paper will discuss the advantages of eco-driving with respect to improvements in fuel consumption as well as pollutant gas emissions. Simulating a conventional passenger vehicle and applying numerical trajectory optimization methods best vehicle operation for a given trip is identified. With hardware-in-the-loop testing on an engine test bench the fuel and emissions are measured. An approach to integrate pollutant emission and dynamically choose the ecologically optimal gear is proposed.
The performance of a vehicle control strategy, in terms of fuel economy improvement and emission reduction, is strongly influenced by driving conditions and drivers' driving styles. The term of 'driving conditions' here means the traffic conditions and road type, which is usually indicated by standard driving cycles, say FTP 75 and NEDC; the term of 'driving styles' here relates to the drivers' behavior, especially how drivers apply pressure on acceleration and brake pedal. To realize optimal fuel economy, it is ideal to obtain the information of future driving conditions and drivers' driving styles. This paper summarizes the methods and parameters that have been utilized to attain this end as well as the results. Based on this study, methods and parameters can be better selected for further improvement of driving conditions prediction and driving style recognition based hybrid electric vehicle control strategy.
Intelligent vehicle systems have introduced the need for designers to consider user preferences so as to make several kinds
of driving features as driver friendly as possible. This requirement raises the problem of how to suitably analyse human performance
so they can be implemented in automatic driving tasks. The framework of the present work is an adaptive cruise control with
stop and go features for use in an urban setting. In such a context, one of the main requirements is to be able to tune the
control strategy to the driver’s style. In order to do this, a number of different drivers were studied through the statistical
analysis of their behaviour while driving. The aim of this analysis is to decide whether it is possible to determine a driver’s
behaviour, what signals are suitable for this task and which parameters can be used to describe a driver’s style. An assignment
procedure is then introduced in order to classify a driver’s behaviour within the stop and go task being considered. Finally,
the findings were analysed subjectively and compared with a statistically objective one.