Publications

Surrogate Safety Measures (SSM) are important for safety performance evaluation, since crashes are rare events and historical crash data does not capture near crashes that are also critical for improving safety. This paper focuses on SSM and their applications, particularly in Connected and Automated Vehicles (CAV) safety modeling. It aims to provide a comprehensive and systematic review of significant SSM studies, identify limitations and opportunities for future SSM and CAV research, and assist researchers and practitioners with choosing the most appropriate SSM for safety studies. The behaviors of CAV can be very different from those of Human-Driven Vehicles (HDV). Even among CAV with different automation/connectivity levels, their behaviors are likely to differ. Also, the behaviors of HDV can change in response to the existence of CAV in mixed autonomy traffic. Simulation by far is the most viable solution to model CAV safety. However, it is questionable whether conventional SSM can be applied to modeling CAV safety based on simulation results due to the lack of sophisticated simulation tools that can accurately model CAV behaviors and SSM that can take CAV’s powerful sensing and path prediction and planning capabilities into crash risk modeling, although some researchers suggested that proper simulation model calibration can be helpful to address these issues. A number of critical questions related to SSM for CAV safety research are also identified and discussed, including SSM for CAV trajectory optimization, SSM for individual vehicles and vehicle platoon, and CAV as a new data source for developing SSM.

Traditional statistical crash prediction models oftentimes suffer from poor data quality and require large amount of historical data. In this paper, we propose a crash prediction method based on a bivariate extreme value theory (EVT) framework, considering both drivers’ perception-reaction failure and failure to proper evasive actions. An unmanned aerial vehicle (UAV) was utilized to collect videos of ten intersections in Fengxian, China, at representative time periods. High-resolution vehicle trajectory data were extracted by a Kanade-Lucas-Tomasi (KLT) technique, based on four detailed metrics were derived including Time-to-accident (TA), Post-encroachment Time (PET), minimum Time-to-collision (mTTC), and Maximum Deceleration Rate (MaxD). TA was expected to capture the chance of perception-reaction failure, while other three metrics were used to measure the probability of failure to proper evasive actions. Univariate EVT models were applied to obtain marginal crash probability based on each metric. Bivariate EVT models were developed to obtain joint crash probability based on three pairs: TA and mTTC, TA and PET, and TA and MaxD. Thus, union crash probability within observation periods can be derived and the annual crash frequency of each intersection was predicted. The predictions were compared to actual annual crash frequencies, using multiple tests. The findings are three-folds: 1. The best conflict metrics for angle and rear-end crash predictions were different; 2. Bivariate EVT models were found to be superior to univariate models, regarding both angle and rear-end crash predictions; 3. TA appeared to be an important conflict metric that should be considered in a bivariate EVT model framework. In general, the proposed method can be considered as a promising tool for safety evaluation, when crash data are limited.

This paper proposes a combined usage of microscopic traffic simulation and Extreme Value Theory (EVT) for safety evaluation. Ten urban intersections in Fengxian District in Shanghai were selected in the study and three calibration strategies were applied to develop simulation models for each intersection: a base strategy with fundamental data input, a semi-calibration strategy adjusting driver behavior parameters based on Measures of Effectiveness (MOE), and a full-calibration strategy altering driver behavior parameters by both MOE and Measures of Safety (MOS). SSAM was used to extract simulated conflict data from vehicle trajectory files from VISSIM and video-based data collection was introduced to assist trained observers to collect field conflict data. EVT-based methods were then employed to model both simulated/field conflict data and derive the Estimated Annual Crash Frequency (EACF), used as Surrogate Safety Measures (SSM). PET was used for EVT measurement for three conflict types: crossing, rear-end, and lane change. EACFs based on three simulation calibration strategies were compared with field-based EACF, conventional SSM based on Traffic Conflict Techniques (TCT), and actual crash frequency, in terms of direct correlation, rank correlation, and prediction accuracy. The results showed that, MOS should be considered during simulation model calibration and EACF based on the full-calibration strategy appeared to be a better choice for simulation-based safety evaluation, compared to other candidate safety measures. In general, the combined usage of microscopic traffic simulation and EVT is a promising tool for safety evaluation.

Nowadays, intelligent transportation system (ITS) planning has been often integrated into transportation planning stage. As a component of ITS, traffic enforcement cameras have been found to reduce dangerous behaviors, such as red-light running and speeding. However, with limited resource, it is important to understand the effects of enforcement cameras on macro-level safety, so that traffic policy-makers can better allocate those resources to improve traffic safety from the planning stage. In this paper, we examined the effects of various traffic enforcement cameras on regional traffic crash risk, considering their interactions with roadway and land use characteristics. The Kunshan city in Suzhou, China was selected in this study and a spatial modeling analysis was applied. According to the modeling results, several conclusions can be drawn: 1. Interaction effects on regional injury/PDO crash risk were found between traffic enforcement cameras and roadway/land use factors; 2. Traffic enforcement cameras were found to be associated with decreased regional crash risk. Among them, red-light running and speeding cameras were associated with the reduction of injury/PDO crash frequency, which can be further enhanced when being installed in certain area (e.g. industrial, commercial, residential land use) and on certain roadways (e.g. major arterials, local roads). Illegal lane changing cameras were associated with the decrease in PDO crash frequency, while such effect on reducing injury crashes was only found as significant on major arterials; 3. The main effects of certain land use and roadway factors appeared to be mediated by traffic enforcement interaction terms. For example, the main effect of industrialized land use was found as insignificant, while the interaction term between industrial area and speeding cameras showed a significant effect of reducing injury/PDO crash frequency. Based on those findings, traffic enforcement cameras, as one of the major components of ITS, need to be carefully considered at the transportation planning stage. In general, this study provides valuable information for policy-makers and transportation planners to improve regional traffic safety, by properly allocating traffic enforcement resources.

It is unknown that how safety knowledge and psychological factors affect e-bike rider safety. Also, group violation behaviors, which are commonly observed in the field, have been rarely examined for e-bike riders in China. In this paper, the effects of safety knowledge and psychological factors on self-reported risky driving behaviors including group violations were examined. A questionnaire was developed to acquire information of e-bike riders in Guilin, Guangxi Province. Explanatory factor analysis was used to examine the reliability of the questionnaire and exclude redundant measurement items. Then, a Structure Equation Model was developed to examine the relationships among safety knowledge, safety attitude, risk perception and risky driving behaviors. After that, multiple regression models were fitted to examine the effect of safety knowledge on each type of risky driving behavior, as well as factors associated with group violations. At last, ANOVA tests were conducted to identify significant differences among e-bike rider groups in safety knowledge, safety attitude, and risk perception. Safety knowledge was found significantly associated with risky driving behaviors for e-bike riders in China, including aggressive driving, erroneous driving, and group violations. E-bike riders severely lack safety knowledge, especially that of traffic rules, including unmarried riders, under-educated riders, riders without driver’s license, younger riders, and riders with little riding experience. Group violations were largely found among e-bike riders, and to be associated with safety knowledge of traffic rules, risk-taking attitude, and riding experience of e-bike riders. The findings could add some new safety implication and be beneficial for developing safety policies and interventions for e-bikes.

Crash risk measures (CRMs) are widely used in safety analysis to complement crash reports. However, none of the existing CRMs are specifically developed for modeling the risk of single-vehicle run-off-road (SVROR) crashes, especially those on horizontal curves. This paper proposes a novel crash risk measure for modeling SVROR crash risk using connected vehicle data. The proposed SVROR crash risk measure (SVROR-CRM) is based on the concept of tetraquark in particle physics. It utilizes the adjusted position deviation risk force and adjusted attitude deviation risk moment to quantify SVROR crash risk. The SVROR crash risk is then estimated by the joint probability using a peak-over threshold approach. The risk threshold is automatically determined via a mean absolute error function. The SVROR-CRM is validated using connected vehicle and crash data from sixteen curves on Interstate 80 in Wyoming. The results suggest that the estimated SVROR crash risks well match historical crash records. Also, it is found that attitude deviation poses a higher risk of SVROR crash than position deviation on horizontal curves. Therefore, it is critical for drivers to steer properly on curves to minimize SVROR crash risks. The proposed approach bridges an important gap in crash risk measure research and can be used to estimate SVROR crash risk and identify unsafe trajectories and high-crash locations and/or periods on highway horizontal curves.

Vehicle detection from surveillance videos is of great significance for various Intelligent Transportation System (ITS) applications. However, existing deep learning methods oftentimes fail under nighttime conditions on account of the lack of sufficient labeled nighttime data. To fill this gap, this paper proposes a novel framework for all-day vehicle detection, by introducing an illumination-adjustable GAN (IA-GAN). The IA-GAN transforms labeled daytime images into multiple nighttime images with diverse illumination, using an adjustable illumination vector as input. Notably, we utilize gray histogram distributions to automatically generate illumination labels, by which IA-GAN gains the knowledge of simulating lights. Following that, we construct a large dataset containing both labeled daytime images and all generated synthetic nighttime images with bounding box labels. Finally, a detector named Day-Night Balanced EfficientDet (DNBED) is developed for all-day vehicle detection. The experiments show that the proposed framework yields promising performance for all-day vehicle detection and competitive results for nighttime vehicle detection compared to existing GANs, indicating the effectiveness of proposed framework. The privacy-sanitized image data and its corresponding labels will be made publicly available at https://github.com/vvgoder/SEU_PML_Dataset.

Pedestrian crossing intention prediction could effectively prevent traffic injuries and improve pedestrian safety. This paper focuses on pedestrian crossing intention prediction from surveillance cameras, which could provide over-the-horizon safety warnings and has the potential to better ensure pedestrian safety, compared with that from on-board cameras. However, most prediction-based methods are designed with a fundamental assumption that the visual data is collected from an on-board camera rather than a bird-eye-view one, thus the prevalent methods in this research domain do not match surveillance scenarios. To deal with this issue, an automated learning framework is proposed, in which a pedestrian-centric environment graph is primarily constructed to reflect visual variations and spatiotemporal relationships between pedestrians and their surroundings. After that, a Graph Convolutional Network (GCN) based environment encoder and a pedestrian-state encoder are designed to extract prominent environment features and pedestrian behavior features, respectively. Finally, an intention prediction decoder is developed to extrapolate the probability of crossing intention. Experimental results demonstrate that each component in the framework contributes to performance improvement and their combination obtains state-of-the-art performance, suggesting the effectiveness and superiority of our framework.

Monitoring-based traffic participant detection (TPD) is a highly desirable but challenging task. So far, deep learning-based methods have attained significant improvements on the TPD task, but oftentimes fail in urban mixed traffic due to the lack of relevant datasets and suitable detectors. In this study, we propose a large and detailed dataset named SEU_PML specialized for monitoring-based TPD in urban mixed traffic. This dataset contains a total of 270,684 objects annotated with 2D bounding box and covers 13 sub-categories, having (i) high-resolution images (from 1920×1080 to 4096×2160 pixels), (ii) high-quality annotation (annotation accuracy reaches 98%), and (iii) rich traffic scenarios covering diverse traffic scenes as well as different weather and illumination conditions. The mixed traffic along with high-quality annotation bring about a variety of small objects. To further address the issue on small object detection, we propose a novel detector named YOLO SOD, which embeds a super-resolution feature extraction module and uses knowledge distillation to learn the knowledge how the detector with high-resolution inputs perceives small objects. Moreover, a novel loss function named S-IoU is designed to enable YOLO SOD to focus more on small objects. Experimental results show that (1) the YOLO SOD detector has an increased mAP of 1.58% and operates approximately four times faster when compared to a state-of-art detector; (2) the detectors trained on the SEU_PML dataset have a strong transferability and could be well applied to traffic participant detection in urban mixed traffic. Our dataset is now available at https://github.com/vvgoder/SEU_PML_Dataset .

Crash detection is of great significance for traffic emergency management. Video-based approaches can effectively save the manpower monitoring cost and have achieved promising results in recent studies. However, they sometimes fail to correctly identify crashes in congested traffic. To fill the gap, this paper proposes a novel appearance-motion network for improving video-based crash detection performance in congested traffic. The appearance-motion network utilizes two paralleled convolutional networks (i.e., an appearance network and a motion network) to extract both appearance features and motion features of crashes. To learn discriminative appearance features for differentiating crashes in congested traffic scene (CCT) with non-crashes in congested traffic scene (NCCT), an auxiliary network combined with a triplet loss are introduced to train the appearance network. To better capture crash motion features in congested traffic, an optical flow learner is built in the motion network and trained to extract more fine-grained motion information. Moreover, a temporal attention module is applied to enable the motion network to focus on valuable frames. Experimental results show that the proposed network achieves a state-of-the-art result on crash detection and the introduction of the three components (i.e., the auxiliary network, the optical flow learner and the temporal attention module) effectively reduces false alarm rate by 28.07% and miss rate by 27.08% on crash detection in congested traffic. Our dataset will be available at https://github.com/vvgoder/Dataset_for_crashdetection.

Traffic equipment detection from surveillance videos is of practical significance for temporary traffic element update in high-precision maps. However, there is little relative research developed due to limited labeled data. Based on a detector dubbed Faster R-CNN, we propose an automated learning framework that utilizes easy-to-obtain Internet images containing traffic equipment to acquire the capability of detecting traffic equipment from surveillance videos. In this framework, an appearance weighting module using a comprehensive feature aggregation method is designed to allow Faster R-CNN to converge and generalize quickly by taking limited data (i.e., less than 30 images per class) as input. To further address the cross-domain issue brought by the domain gap between the Internet images and the surveillance video frames, a domain adaptation learning scheme is developed, which aims to align the two domains and guide the framework to learn more robust domain-invariant features. Experimental results show that both the appearance weighting module and the domain adaptation learning scheme could bring a great performance improvement. Moreover, the combination of the two results in a state-of-the-art performance (mAP of 44.6%) even if only 30 training images per class are provided. To sum up, the proposed framework is suitable for traffic equipment detection from surveillance videos and provides an inspiration for other detection tasks with limited and cross-domain data, allowing humans to reduce their efforts and time required for arduous data collection and annotation.

Traffic equipment detection from surveillance videos is of practical significance for temporary traffic element update in high-precision maps. However, there is little relative research developed due to limited labeled data. Based on a detector dubbed Faster R-CNN, we propose an automated learning framework that utilizes easy-to-obtain Internet images containing traffic equipment to acquire the capability of detecting traffic equipment from surveillance videos. In this framework, an appearance weighting module using a comprehensive feature aggregation method is designed to allow Faster R-CNN to converge and generalize quickly by taking limited data (i.e., less than 30 images per class) as input. To further address the cross-domain issue brought by the domain gap between the Internet images and the surveillance video frames, a domain adaptation learning scheme is developed, which aims to align the two domains and guide the framework to learn more robust domain-invariant features. Experimental results show that both the appearance weighting module and the domain adaptation learning scheme could bring a great performance improvement. Moreover, the combination of the two results in a state-of-the-art performance (mAP of 44.6%) even if only 30 training images per class are provided. To sum up, the proposed framework is suitable for traffic equipment detection from surveillance videos and provides an inspiration for other detection tasks with limited and cross-domain data, allowing humans to reduce their efforts and time required for arduous data collection and annotation.

Deep learning methods have attained promising performance on road defect detection from on-board cameras. However, they oftentimes rely heavily on well-annotated datasets with sufficient samples, limiting the practical applications when only few labeled samples are available. To fill this gap, this paper proposes a framework based on Faster Region-Convolutional Neural Network (Faster R-CNN) for road defect detection with scarce and cross-domain data. First, a defect weighting branch is developed to enable Faster R-CNN to quickly learn to detect road defects with few annotated data, then a data augmentation method is proposed to enlarge the abundance of annotated data and alleviate the cross-domain issue. Experimental results demonstrate that the proposed framework has attained better performance compared to a state-of-the-art few-shot detector, in terms of an improved mean average precision of 1.83% when only limited samples (i.e., 30 images per category) are provided for training. In the future, the proposed framework could also be extended to other detection tasks with limited data (e.g., construction vehicle detection), allowing humans to reduce their efforts and time required for arduous data collection and annotation.

Automatic waste detection in natural environments exhibits a great potential to improve the efficiency and reduce the labor cost of waste management. Recent deep learning-based waste detectors rely heavily on substantial annotated samples for training, but annotating sufficient samples for various categories of waste is labor-intensive and time-consuming. To address this issue, this paper simulates the visual system of human beings and develops a few-shot waste detection framework. To enable the proposed framework more suitable for waste detection, a waste proposal module using a comprehensive feature fusion manner is designed to allow the features of support images to fully interact with those of query images, guiding the framework to generate more potential region proposals containing waste. Also, a waste classification module using soft attention mechanism and foreground mask is designed to alleviate the issue of spatial misalignment and achieve the fine-grained classification towards waste-related proposals. The proposed framework is a general detection framework which can flexibly detect various categories of waste with few labeled samples (i.e., less than 30 instances per category). Experimental results show that the proposed framework achieves a mean average precision of 31.16% over 12 waste categories when only few samples (i.e., 30 instances per category) are provided, surpassing a state-of-the-art few-shot detector named AFDNet by 1.68%. This data scale-insensitive nature allows humans to reduce the effort and time required for laborious waste image collection and annotation, significantly increasing the flexibility of automatic waste detection and boosting the efficiency of waste management.

Extensive research has examined the potential benefits of Automated Vehicles (AVs) for increasing traffic capacity and improving safety. However, previous studies on AV longitudinal control have focused primarily on control stability and instability or tradeoffs between safety and stability, neglecting the importance of vehicle damping characteristics. This study aims to demonstrate the significance of explicitly considering safety in addition to stability in AV longitudinal control through damping behavior analysis. Specifically, it proposes a safety-oriented AV longitudinal control and provides recommendations on the control parameters. For the proposed AV control, an Adaptive Cruise Control (ACC) model is integrated with damping behavior analysis to model AV safety under continuous traffic perturbations. Numerical simulations are conducted to quantify the relationship between mobility and safety for AVs considering both damping behavior and control stability. Different ACC control parameters are evaluated in terms of damping and stability properties, and their safety impacts are assessed based on various surrogate safety measures such as Deceleration Rate to Avoid Crash (DRAC), Crash Potential Index (CPI) and Time-Integrated Time-to-collision (TIT). The results indicate that an underdamped state (ACC damping ratio < 1) is less safe than the critically damped state (ACC damping ratio = 1) and the overdamped state (ACC damping ratio > 1). Furthermore, given the same AV car-following time lag, ACC with a damping ratio between 1 and 1.2 provides better safety performance. Increasing the AV car-following time lag can improve both safety and stability when the remaining ACC control parameters are kept the same. In this case, the optimal safety-oriented ACC regions also increase. The findings of this study provide important insights into designing safe and stable AV longitudinal control algorithms.

The concepts of Connected and Automated Vehicles (CAV) and vehicle platooning have generated high expectations regarding the safety performance of future transportation systems. Existing CAV longitudinal control research primarily focuses on efficiency and control stability, by considering different inter-vehicle spacing policies. In very few cases, safety was also considered as a constraint, but not in the main control objectives. Theoretically, stability can only guarantee that CAV platoons eventually achieve an equilibrium state but is unable to promise safety along the process of achieving equilibrium. It is important to note that CAV does not mean absolutely safe, and its longitudinal or platoon control safety performance depends on how the control algorithms are designed, how accurately it can detect and predict its lead vehicle’s (could be a human-driven vehicle) next move, and other practical factors such as control and communication delays. To optimize CAV platoon safety, this study integrates surrogate safety measures (SSM) and model predictive control (MPC) into CAV longitudinal control for trajectory optimization. SSM has been widely adopted for modeling the safety consequences of various vehicle control strategies and identifying near-crash events from either simulated or field-captured traffic data. This study directly incorporates three typical SSM into the longitudinal control objectives of CAV and constructs a state-space MPC algorithm to model how these SSM vary as a result of CAV dynamics. Numerical examples are provided to show the performance of these SSM-based optimal CAV longitudinal control methods under traffic flow perturbations. To further confirm the necessity of explicitly considering SSM in CAV longitudinal control and its effectiveness in reducing rear-end collision risk, the proposed methods are compared with three classical longitudinal control models that do not consider SSM based on microscopic traffic simulation. It is noted that all SSM-based optimal control methods perform better than others as manifested by some key risk indicators, demonstrating the importance of explicitly considering SSM and safety in CAV longitudinal control.

Automatic fruit blossom detection plays a crucial role in agricultural intelligence to predict fruit yield. Existing deep learning methods for vision-based fruit blossom detection oftentimes rely on large labeled samples and are tailor for a single category of fruit blossoms, limiting their flexibility and adaptability in real applications. In this paper, a fast and data-efficient framework is proposed to achieve multi-class fruit blossom detection using few labeled samples, which not only enhances the flexibility of model training by reducing the need for extensive samples but also extends the model’s applicability to a wider range of blossom categories. Specifically, the proposed framework incorporates the paradigm of few-shot object detection into a lightweight two-stage detector named CenterNet2. To improve the recall of blossom proposals, we introduce a location guidance module (LGM) to highlight foreground regions in query images that bear resemblance to the blossom reference provided by support images. Moreover, a contrastive learning scheme (CLS) is introduced to further distinguish fruit blossom categories with similar appearances, enhancing classification accuracy. Experimental results demonstrated that our framework achieved comparable performance to existing state-of-the-art models, with a mean average precision of 74.33% over 12 categories of fruit blossoms and a frame per second rate of 47. In summary, our proposed framework can reduce data dependency and enhance flexibility of existing deep learning detectors, making it a promising solution for automatic monitoring in real-world agricultural applications.

Abnormal trajectory detection from surveillance cameras is a highly desirable but challenging task, especially for online traffic incident alert on freeways. Existing methods are mainly customized for offline alert and easily suffer from false alerts when applying them to online alert. To fill this gap, an anomaly trajectory detection framework is proposed for online traffic incident alert on freeways. Based on a LSTM autoencoder, this framework introduces an adversarial learner (AL) for offline training and an abnormal trajectory discriminator (ATD) for online alert. The adversarial learner uses an additional adversarial loss to enable the autoencoder to learn a better normal trajectory pattern that is beneficial for reducing false alerts, while an abnormal trajectory discriminator is established and trained to detect small mean shift and filter out instantaneous false alerts. The experimental results show that our proposed framework effectively filters out false alerts and obtains a state-of-art performance (AUC = 0.97) compared to existing methods. Moreover, our framework could timely alert traffic incidents within 0.25 s, which is significant for timely preventing the occurrence of traffic crashes and improving the response speed of incident management on freeways.

Predicting future traffic state (e.g., traffic speed, volume, travel time, etc.) accurately is highly desirable for traffic management and control. However, network-wide traffic flow has complicated spatial-temporal dependencies, making it challenging to predict. This study proposes a multi-weighted graph 3D convolution network (MWG3D) to predict future network-wide traffic speed, considering the spatial-temporal heterogeneous effects of multiple external factors (i.e., points of interests (POIs), roadway physical characteristics and incidents). The network is composed of a Graph-3D convolution (G3D) module and an incident impact module. In G3D module, a weighted graph convolution is developed first, which extracts complex spatial dependencies of traffic flow considering heterogeneous effects of POIs and roadway physical characteristics. These external factors have great influence on the periodicity of human daily activities, which in turn cyclically affect traffic flow. The weighted graph convolution is further connected with 3D convolutions to extract temporal dependencies of traffic flow, accounting for temporal heterogeneous effects of these external factors. An incident impact module is separately developed to account for spatial-temporal heterogeneous effects of incidents. These external factors could lead to abrupt and temporary changes in traffic flow. The proposed network is evaluated on two real-world datasets. The results show that MWG3D outperforms a selection of the state-of-the-art models. Furthermore, the spatial-temporal heterogeneous effects of external factors are crucial to prediction accuracy.

Abnormal trajectory detection from surveillance cameras is a highly desirable but challenging task, especially for online traffic incident alert on freeways. Existing methods are mainly customized for offline alert and easily suffer from false alerts when applying them to online alert. To fill this gap, an anomaly trajectory detection framework is proposed for online traffic incident alert on freeways. Based on a LSTM autoencoder, this framework introduces an adversarial learner (AL) for offline training and an abnormal trajectory discriminator (ATD) for online alert. The adversarial learner uses an additional adversarial loss to enable the autoencoder to learn a better normal trajectory pattern that is beneficial for reducing false alerts, while an abnormal trajectory discriminator is established and trained to detect small mean shift and filter out instantaneous false alerts. The experimental results show that our proposed framework effectively filters out false alerts and obtains a state-of-art performance (AUC = 0.97) compared to existing methods. Moreover, our framework could timely alert traffic incidents within 0.25 s, which is significant for timely preventing the occurrence of traffic crashes and improving the response speed of incident management on freeways.

The individual choices of parkers, based on self-interests, can result in the inefficient use of on-street parking-space resources as well as chaotic parking conditions (e.g., secondary path interweaving). Therefore, an ensemble optimization strategy for dynamic parking-space allocation (EO-DPSA) is proposed for systematically optimizing on-street parking-space use. First, the issue of inefficient parking-space use caused by the individual choice of parkers is investigated. Subsequently, the EO-DPSA technique is developed for optimizing parking-space use from a system level by improving the parking sequence for eliminating parking-entry conflicts. Consequently, the compliance rate of parking guidance is increased, and secondary path interweaving is prevented. Next, the optimization effect of the EO-DPSA technique is validated based on the smart parking data obtained from Guilin City. The obtained results indicate that after optimization via the EO-DPSA method, the parking-space use of 81 parking areas is optimized. Hence, 2.86–66.67% of redundant parking spaces can be discarded in 41 parking areas. In particular, a parking compliance rate of 99.023% is achieved. In addition, 97.82% of the parking phases exhibit better results in avoiding secondary path interweaving after optimization

The development of surrogate safety measures is essential due to the problems of availability and quality of historical crash data. The Aggregate Conflict Propensity Metric (ACPM) is a surrogate metric recently proposed and it is based on conflict studies and traffic simulations. ACPM is expected to be capable of assessing the relative safety levels of traffic facilities and/or treatments in order to help traffic engineers to select appropriate treatments based on traffic safety estimates. This paper presents three experimental tests conducted to evaluate the reliability of ACPM. In each test, ACPM is compared to a traditional conflict indicator in terms of identifying and ranking safety of traffic conditions under various traffic volumes based on traffic simulations. ACPM shows its strength and reliability in all three tests, as it provides results highly consistent with the Highway Safety Manual. The experimental tests indicate that ACPM is a promising surrogate safety measure that can appropriately identify relative safety among traffic treatments and/or facilities and provide traffic engineers with useful information on potential safety impact.

The Conflict Propensity Metric (CPM) and the Aggregate Conflict Propensity Metric (ACPM) are two simulation-based conflict metrics recently proposed as surrogate safety measures. The two metrics are derived through a stochastic process incorporating distributions of driver reaction time (RT) and vehicle maximum braking rates (MABR). This paper presents sensitivity analyses on the two metrics, by altering the parameters (i.e. mean and standard deviation) of RT distributions. Both RT mean and standard deviation affect the estimates of CPM for the three conflict types examined here (i.e. crossing, rear-end and lane change), and the impacts vary by conflict types, indicating the need of carefully evaluating and considering both RT distributions and conflict types when developing simulation-based conflict metrics. A sensitivity analysis based on field data showed that different RT distributions have an impact on ACPM and could affect the reliability of ACPM in identifying relative safety and its correlation with actual crashes. The analysis here identified the potential existence of different “realistic” RT distributions for different conflict types and the suggested values are considered reasonable or consistent with prior findings. ACPM has proved to have potential to further improve its accuracy using more suitable RT distributions. However, dedicated RT distributions for specific conflict types are lacking, especially a joint RT distribution conditional various factors, impeding the improvement of ACPM. In general, sensitivity analyses have shown the strength of the process of deriving CPM and ACPM by providing reasonable findings, as well as pointing out interesting and important future research directions, such as finding actual RT distributions for different conflict types.

The Conflict Propensity Metric (CPM) and the Aggregate Conflict Propensity Metric (ACPM) are two simulation-based conflict metrics recently proposed as surrogate safety measures. The two metrics are derived through a stochastic process incorporating distributions of driver reaction time (RT) and vehicle maximum braking rates (MABR). This paper presents sensitivity analyses on the two metrics, by altering the parameters (i.e. mean and standard deviation) of RT distributions. Both RT mean and standard deviation affect the estimates of CPM for the three conflict types examined here (i.e. crossing, rear-end and lane change), and the impacts vary by conflict types, indicating the need of carefully evaluating and considering both RT distributions and conflict types when developing simulation-based conflict metrics. A sensitivity analysis based on field data showed that different RT distributions have an impact on ACPM and could affect the reliability of ACPM in identifying relative safety and its correlation with actual crashes. The analysis here identified the potential existence of different “realistic” RT distributions for different conflict types and the suggested values are considered reasonable or consistent with prior findings. ACPM has proved to have potential to further improve its accuracy using more suitable RT distributions. However, dedicated RT distributions for specific conflict types are lacking, especially a joint RT distribution conditional various factors, impeding the improvement of ACPM. In general, sensitivity analyses have shown the strength of the process of deriving CPM and ACPM by providing reasonable findings, as well as pointing out interesting and important future research directions, such as finding actual RT distributions for different conflict types.

Existing crack recognition methods based on deep learning often face difficulties when detecting cracks in complex scenes such as brake marks, water marks, and shadows. The inadequate amount of available data can be primarily attributed to this factor. To address this issue, a controllable generative model of pavement cracks is proposed that can generate crack images in complex scenes by leveraging background images and crack mask images. The proposed model, the crack diffusion model (CDM), is based on the diffusion model network, which enables better control over the position and morphology of cracks by adjusting the conditional input of cracks. Experiments show that CDM has several advantages, including high definition, controllability, and sensitivity to narrow cracks. Utilizing CDM to create a synthetic crack data set in complex scenes resulted in substantial improvements of crack detection and segmentation. The method proposed in this study can effectively alleviate the effort required for data acquisition and labeling, especially in complex scenes.

Three-year crash data are retrieved from the crash reports (2012–2014) from the Taixing Police Department. A mixed logit model is introduced to explore significant factors associated with fault assignment, as well as accounting for similarity among fault assignment and heterogeneity within unobserved variables. The modeling results indicate some interesting new findings. First, precrash behaviors of both drivers and e-bike riders are found to be significant to fault assignment. Second, bike lane and median type are significantly associated with e-bike rider fault commitment. Third, specific groups of e-bike riders (low-educated and older) and drivers (heavy good vehicles) are more likely to be at fault in e-bike crashes. Last, crash location and the built environment have significant correlations with faulty behaviors of e-bike riders. Safety countermeasures are proposed including (1) the deployment of traffic design and control elements including physically separated bike lanes, medians, video surveillance systems for e-bike riders, and left-turning treatments for nonmotorists (e.g., a 2-step e-bike left turning); (2) the amendment of the current traffic regulations on drunk e-bike riders and child e-bike passengers; (3) the development of a license system for specific e-bike rider groups (older and low-educated) and a safety campaign for drivers (to increase safety awareness when parking on-street or driving heavy good vehicles). Some interesting future research topics are also suggested: e-bike riders' behaviors at unsignalized intersections and mid-block openings, e-bike safety in suburban areas, and an in-depth study of the effect of the built environment on e-bike safety.

Introduction: Traffic crashes could result in severe outcomes such as injuries and deaths. Thus, understanding factors associated with crash severity is of practical importance. Few studies have deeply examined how prior violation and crash experience of drivers and roadways are associated with crash severity. Method: In this study, a set of risk indicators of road users and roadways were developed based on their prior violation and crash records (e.g., cumulative crash frequency of a roadway), in order to reflect certain aspect or degree of their driving risk. To explore the impacts of those indicators on crash severity and complex interactions among all contributing factors, a Bayesian network approach was developed, based on citywide crash data collected in Kunshan, China from 2016 to 2018. A variable selection procedure based on Information Value (IV) was developed to identify significant variables, and the Bayesian network was employed to explicitly explore statistical associations between crash severity and significant variables. Results: In terms of balanced accuracy and AUCs, the proposed approach performed reasonably well. Bayesian modeling results indicated that the prior crash/violation experiences of road users and roadways were very important risk indicators. For example, migrant workers tend to have high injury risk due to their dangerous violation behaviors, such as retrograding, red-light running, and right-of-way violation. Furthermore, results showed that certain variable combinations had enhanced impacts on severity outcome than single variables. For example, when a migrant worker and a non-motorized vehicle are involved in a crash happening on a local road with high cumulative violation frequency in the previous year, the probability for drivers suffering serious injury or fatality is much higher than that caused by any single factor. Practical applications: The proposed methodology and modeling results provide insights for developing effective countermeasures to reduce crash severity and improve traffic system safety performance.

This study aimed to investigate the characteristics and patterns of the connected and autonomous vehicle (CAV) involved crashes. Method: The crash data were collected from the reports of CAV involved crash submitted to the California Department of Motor Vehicles. The descriptive statistics analysis was employed to investigate the characteristics of CAV involved crashes in terms of crash location, weather conditions, driving mode, vehicle movement before crash occurrence, vehicle speed, collision type, crash severity, and vehicle damage locations. The bootstrap based binary logistic regressions were then developed to investigate the factors contributing to the collision type and severity of CAV involved crashes. Results: The results suggested that the CAV driving mode, collision location, roadside parking, rear-end collision, and one-way road are the main factors contributing to the severity level of CAV involved crashes. The CAV driving mode, CAV stopped or not, CAV turning or not, normal vehicle turning or not, and normal vehicle overtaking or not are the factors affecting the collision type of CAV involved crashes.

Determining spatiotemporal impact areas of incidents plays a significant role in incident impact analysis. Although existing empirical methods have proven to be promising, they suffer from the drawbacks that limit their wide applications in automated freeway safety management. This study presents a data‐driven approach to automatically determining the spatiotemporal impact areas of freeway incidents. The spatiotemporal contour plots were first constructed using three representative traffic measures. Next, a nonrecurrent congestion area identification method based on fuzzy clustering was developed. To distinguish possible multiple independent blocks in the nonrecurrent congestion area, a clustering algorithm based on graph theory was adopted. The incident impact areas were then determined by conducting a postprocessing strategy. The incident records and the associated traffic flow data, collected on I‐5 freeway segments in San Diego Region, CA, were used to evaluate the proposed approach. Experimental results show the proposed approach can automatically and properly determine incident impact areas while accounting for the uncertainty resulting from traffic variations.

Car-following models have been widely applied and made remarkable achievements in traffic engineering. However, the traffic micro-simulation accuracy of car-following models in a platoon level, especially during traffic oscillations, still needs to be enhanced. Rather than using traditional individual car-following models, we proposed a new trajectory generation approach to generate platoon level trajectories given the first leading vehicle’s trajectory. In this article, we discussed the temporal and spatial error propagation issue for the traditional approach by a car following block diagram representation. Based on the analysis, we pointed out that error comes from the training method and the model structure. In order to fix that, we adopt two improvements on the basis of the traditional LSTM-based car-following model. We utilized a scheduled sampling technique during the training process to solve the error propagation in the temporal dimension. Furthermore, we developed a unidirectional interconnected LSTM model structure to extract trajectories features from the perspective of the platoon. As indicated by the systematic empirical experiments, the proposed novel structure could efficiently reduce the temporal-spatial error propagation. Compared with the traditional LSTM-based car-following model, the proposed model has almost 40% less error. The findings will benefit the design and analysis of micro-simulation for platoon-level car-following models.