Using traffic conflicts to evaluate real-time vehicle-level crash risk is promising for proactive traffic safety management. Previous studies have typically employed pre-trained deep learning models based on microscopic trajectory information to predict conflicts. However, such models often suffer from limited transferability, as they fail to capture the dynamic effects of safety factors and neglect macroscopic traffic flow characteristics. To address these limitations, this study develops a retrieval-augmented framework integrating macroscopic traffic flow and microscopic vehicle trajectory, which comprises two sequential components: a Retrieval-Augmented Multilayer Perceptron model to identify conflict-prone lanes using traffic flow data; and a Retrieval-Augmented Convolutional Neural Network-Long Short-Term Memory model to screen for conflict-involved vehicles in high-risk lanes using trajectory data. Results show: (1) the retrieval-augmented module significantly improves prediction accuracy with 22.5% higher precision and 11.0% higher sensitivity; (2) incorporating traffic flow characteristics enhances the precision and sensitivity of vehicle-specific conflict prediction by 6.3% and 7.6%, respectively; (3) the advantages of the retrieval-augmented module are more pronounced for data-constrained scenarios. This study demonstrates the potential of retrieval-augmented techniques to enhance deep learning-based real-time conflict prediction models. The modeling framework provides valuable insights for practical application in proactive safety warning systems, especially in Vehicle-to-Infrastructure communication environments.

Lane changing is a crucial maneuver for connected and autonomous vehicles (CAV) to achieve efficient and safe driving, demanding tightly coupled decision-making and trajectory planning supported by accurate predictions of surrounding traffic. We introduce an integrated architecture for CAV coupling three core functionalities: predictive modeling of adjacent vehicle trajectories, game-theoretic multi-agent decision strategies, and adaptive real-time path generation. The framework comprises three core modules: First, a learning-based model estimates surrounding vehicles' motions. Second, a multi-vehicle game-theoretic decision-making module evaluates various lane change scenarios, such as successful lane change, failed lane-changing attempts, and re-lane change after failed lane-changing attempts. Third, a quintic-poluomial planner continuously plans the CAV's trajectory in real time. We conduct joint simulation experiments for specific lane changing scenarios, in which CAV strictly implements framework functions and evaluate three performance dimensions: forecasting precision, strategic decision efficacy, and trajectory optimization metrics. In mandatory lane change scenarios, the proposed framework achieved an 80 % success rate, with the full process from intention initiation to successful lane-change completion averaging 4.0 s, outperforming benchmark models. In discretionary scenarios, it reached a 70 % success rate and a 6.8-second completion time, compared to 30 % and 9.0 s for existing methods. These results demonstrate that integrating motion prediction, game-theoretic decision making, and adaptive planning enhances both the safety and efficiency of CAV lane change behavior.

Existing studies on trajectory optimization for cooperative automated driving systems (C-ADS) equipped vehicles at signalized intersections operate under a simplified assumption of cooperative behaviour: all vehicles accept and follow to the prescribed plans. To investigate trajectory optimization for C-ADS-equipped vehicles with different cooperation classes, a deep deterministic policy gradient (DDPG) algorithm was developed within a reinforcement learning (RL) framework, alongside baseline implementations of trajectory smoothing (TS)-based C-ADS systems and human-driven vehicle scenarios. Experimental results indicate that the proposed methodology achieves significant reductions in average travel time (53.59%) and stop times, compared to benchmark approaches. Furthermore, novel insights into the performance improvements at signalized intersections were derived from analysing different cooperation classes of C-ADS-equipped vehicles via the RL model, providing critical guidance for refining control strategies in cooperative automated driving systems. This study validates that RL models utilizing the DDPG algorithm serve as effective tools for enhancing the performance of cooperative automated driving systems.

Recently, real-time traffic conflict prediction has drawn increasing attention due to its significant potential in proactive traffic safety systems. While various statistical and machine learning models have been developed for conflict prediction, transferability remains a fundamental issue across these models. Specifically, the predictive performance of a real-time conflict prediction model developed for a specific location can significantly decline when directly applied to a new location without any modifications, primarily due to substantial differences in traffic environments between these areas. To address this gap, this study proposed a novel deep transfer learning approach aimed at enhancing the transferability of real-time conflict prediction models. Initially, a real-time conflict prediction framework was designed utilizing trajectory data for merging areas with consideration of temporal variations in traffic flow characteristics. Subsequently, the Gated-Transformer, Fully Convolutional Networks (FCN), Long Short-Term Memory Fully Convolutional Networks (LSTM-FCN), and Multivariate Long Short-Term Memory Fully Convolutional Networks (MLSTM-FCN) were employed as backbone feature extraction networks to capture the hidden correlations between time-varying traffic flow characteristics and traffic conflicts. After that, an independent transfer learning architecture was established to assess the similarity of the distribution of traffic flow characteristics at different locations, based on the maximum mean discrepancy criteria. For empirical evaluation, merging areas from the exiD dataset were differentiated into source and target domains. The results demonstrated that the Gated-Transformer model outperforms other baseline models (FCN, LSTM-FCN and MLSTM-FCN) in both feature extraction and predictive performance, achieving an F1 score of 0.864 and an area under the curve (AUC) of 0.980. Furthermore, the transfer learning architecture can substantially enhance the predictive performance of a model trained in the source domain when applied to the target domain. In particular, the F1 score and AUC for the Gated-Transformer model improved by 11.9% and 10.2%, respectively, after incorporating the transfer learning architecture. Finally, the optimal values of key model parameters, including the sliding time window (6 s) and the prewarning time (5 s), were recommended for practical applications through sensitivity analysis. This study illustrates the potential of the deep transfer learning approach as a reliable and effective alternative to improve the transferability of real-time conflict prediction models. Additionally, results from this study can offer valuable insights for practical applications in traffic safety warning systems, particularly in vehicle-to-infrastructure traffic environments.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

Real-time vehicle safety prediction is critical in roadway safety management as drivers or vehicles can be altered beforehand to take corresponding evasive actions and avoid possible collisions. This study proposes a physics-informed multi-step real-time conflict-based vehicle safety prediction model to enhance roadway safety. Physics insights (i.e., traffic shockwave properties) are combined with data-driven features extracted from deep-learning techniques to improve prediction accuracy. A time series of future vehicle safety indicators are predicted such that vehicles/drivers have enough time to take precautions. The safety indicator at each time stamp is a continuous value that the sign reflects the presence of conflict risks, and the absolute value indicates the conflict risk level to advise different magnitudes of evasive actions. A customized loss function is developed for the proposed prediction model to give more attention to risky events, which are the focus of safety management. The prediction superiority of the proposed model is proven through numerical experiments by comparing it with three benchmarks constructed based on the literature. Further, sensitivity analysis on key model parameters is carried out to advise parameter selections in developing real-world conflict-based vehicle safety prediction applications.

Random parameters logit models have become an increasingly popular method to investigate crash-injury severities in recent years. However, there remain potential elements of the approach that need clarification including out-of-sample prediction, the calculation of marginal effects, and temporal instability testing. In this study, four models are considered for comparison: a fixed parameters multinomial logit model; a random parameters logit model; a random parameters logit model with heterogeneity in means; and a random parameters logit model with heterogeneity in means and variances. A full simulation of random parameters is undertaken for out-of-sample injury-severity predictions, and the prediction accuracy of the estimated models was assessed. Results indicate, not surprisingly, that the random parameters logit model with heterogeneity in the means and variances outperformed other models in predictive performance. Following this, two alternative methods for computing marginal effects are considered: one using Monte Carlo simulation and the other using individual estimates of random parameters. The empirical results indicate that both methods produced defensible results since the full distributions of random parameters are considered. Finally, two testing alternatives for temporal instability are evaluated: a global test across all time periods being considered, and a pairwise time-period to time-period comparison. It is shown that the pairwise comparison can provide more detailed insights into possible temporal variability. (C) 2021 Elsevier Ltd. All rights reserved.

为准确识别冰雪路面下高速公路事故致因因素及严重程度影响因素,构建了基于故障树和贝叶斯网络的综合模型.在转化后的贝叶斯网络中增设3条有向弧,并根据事故严重程度将叶节点分为3种状态,对叶节点的条件概率表进行更新,基于构建的综合模型进行贝叶斯网络逆向推理和敏感性分析.结果表明:能见度低、不良天气(雨、雪、雾)、货车、夜间无照明、驾驶经验不足、超速行驶、未保持安全距离等是诱发冰雪路面高速公路事故的高风险因素;超载、货车、违法上路等在冰雪路面条件下,更易加重事故严重程度.故障树和贝叶斯网络相结合的方法可为事故因素分析提供新视角.

为揭示频繁换道对交通流的影响,构建一种考虑频繁换道倾向的元胞自动机模型.对NGSIM车辆轨迹数据进行提取、统计分析,研究车辆速度、车头间距对车辆换道倾向的影响;改进换道概率函数刻画上述影响,并定义驾驶行为倾向函数表征频繁换道对目标车道紧邻后车的影响,构建考虑频繁换道倾向的双车道元胞自动机STCA-FLC模型.结果表明:STCA-FLC模型较STCA模型在车流密度小于13辆/km时,车辆平均速度、流量最大分别提高约6.67％、3.02％;在车流密度大于13辆/km时,车辆平均速度、流量最大分别降低约9.48％、14.91％.

This study thoroughly investigated factors affecting crash occurrence using detailed data of crash, traffic condition and freeway geometries. To fully account for heterogeneity induced by unobserved characteristics of crash factors, a mixed logit model with mean-variance heterogeneity was estimated as an alternative to the commonly used mixed logit model and the fixed parameters logit model. Results indicate that the mixed logit model with mean-variance heterogeneity could improve the goodness-of-fit and was more flexible in accounting for unobserved heterogeneity compared with its counterparts. Additionally, by allowing means and variances of random parameters to be estimable functions of explanatory variables, the safety effect of interactions among multiple factors was concluded, for example: (1) sharp curves resulted in an increasing risk of crash and the rate of increase was positively correlated with the distance travelled by vehicles along a steep downgrade; (2) the adverse safety effect of steep downgrade increased with the distance covered by vehicles, especially for segments with high proportion of heavy trucks; (3) downhill segments with steep slopes were particularly dangerous. Findings from this study are expected to provide an insightful knowledge to the mechanism of crash occurrence and should be beneficial to design and manage safer freeways.

Unobserved heterogeneity has been a major challenge in developing reliable road safety models. A number of statistical techniques have been developed to account for unobserved heterogeneity and of which, the random parameters approach is one of the most effective method and has been frequently used recently. In this study, to ascertain the performance of various methods that can account for unobserved heterogeneity, the following six models in the negative binomial framework were estimated and thoroughly evaluated: 1) a fixed parameters model; 2) a random parameters model; 3) a random parameters model with heterogeneity in means; 4) a random parameters model with heterogeneity in means and variances; 5) a correlated random parameters model; 6) a correlated random parameters model with heterogeneity in means. These models were thoroughly evaluated from several angles including statistical fit, predictability, causality, marginal effects, explanatory power and practicality. Results indicate that: 1) the simple fixed parameters model resulted in reduced statistical fit, relatively inaccurate predictions, restrictive inferences and added risk of biased marginal effects, though the model is simple for application and interpretive analysis; 2) introducing random parameters could improve the goodness-of-fit, prediction accuracy and the ability to uncover causality; 3) compared with a basic random parameters model, considering the heterogeneity in means/variances or correlation of random parameters brought further improvement in statistical fit, predictive performance and causal inferences; 4) estimating complex models with heterogeneity in means/variances or correlated random parameters provided more insights, but at the expense of significant increase in computational cost and less intuitive outputs; 5) using global means of random parameters rather than the simulated observation-specific parameters was more likely to result in biased marginal effects and erroneous safety measures; 6) new insights into safety factors and their interactions were derived. Results from this study are expected to provide safety analysts with additional guidance for choosing appropriate models when unobserved heterogeneity exists. Additionally, the concluded interactions of safety factors can potentially help develop more effective safety measures. (C) 2021 Elsevier Ltd. All rights reserved.

Numerous studies have previously used a variety of count-data models to investigate factors that affect the number of crashes over a certain period of time on roadway segments. Unlike past studies which deal with crash frequency, this study views the crash rates directly as a continuous variable left-censored at zero and explores the application of an alternate approach based on tobit regression. To thoroughly investigate the factors affecting freeway crash rates and the potentially temporal instability in the effects of crash factors involving traffic volume, freeway geometries and pavement conditions, a classic uncorrelated random parameters tobit (URPT) model and a correlated random parameters tobit (CRPT) model were estimated, along with a conventional fixed parameters tobit (FPT) model. The analysis revealed a large number of safety factors, including several appealing and interesting factors rarely studied in the past, such as the safety effects of climbing lanes and distance along composite descending grade. The results also showed that the CRPT model was not only able to reflect the heterogeneous effects of various factors, but also able to estimate the underlying interactions among unobserved characteristics, and therefore provide better statistical fit and offer more insights into factors contributing to freeway crashes than its model counterparts. Additionally, the results showed significant temporal instability in CRPT models across the studied time periods indicating that crash factors (including unobserved characteristics and the underlying interactions among them) and their effects on crash rates varied over time, and more attentions should be paid when interpreting crash data-analysis findings and making safety policies. The modeling technique in this study demonstrates the potential of CRPT model as an effective approach to gain new insights into safety factors, particularly when the heterogeneous effects of factors on safety are interactive. Additionally, findings from this study are also expected to assist in developing more effective countermeasures by better understanding the safety effects of factors associated with freeway design characteristics and pavement conditions.

为准确分析公交消费数据不完整情况下的公交出行特征,基于乘客上车刷卡数据、支付宝扫码数据及公交GPS数据,运用时空匹配法和出行链理论挖掘分析乘客上下车站点、公交线路OD矩阵、出行空间分布特性及消费时间分布特征.实际验证结果表明:1)使用IC卡和支付宝的乘客数量近似相等,使用现金人数较少,约占整体的6％;2)乘客出行次数在2次以下占总数的84％,换乘需求较少,公交可达性较高;3)高峰期消费次数均超过25000次/h,约占全天总数的23％,居民出行目的较为单一,大部分往返于居民区与办公商业区,与实际情况相符.

To improve the traffic safety at the two-phase intersection, collision risk value between the left-turn vehicle and the straight vehicle in the opposite direction could be used to judge their order. Based on the safety and risk, this method analyzed the process of conflict, modified the operating parameters and then collision risk analysis model was proposed. In the paper, one practical example was given. The result shows that the method can be applied to design the signal phasing-sequence of traffic at signalized intersection, which not only improve safety theory but also provide the theoretical basis for security information service.

Unobserved heterogeneity induced by omitted variables is a major challenge in developing reliable road safety models. In recent years, the random parameters negative binomial (RPNB) model has been used frequently in crash frequency analysis to account for unobserved heterogeneity. However, the majority of past studies of the RPNB model assumed that there was no correlation between different sources of unobserved heterogeneity, which is not always true given the complex interactions of safety factors. Compared with the RPNB model, a more flexible random parameters model that is the correlated random parameters negative binomial with heterogeneity in means (CRPNBHM) model was proposed in this study. Results indicate that the CRPNBHM model could not only capture the otherwise unobserved heterogeneity, but also track the underlying correlation among different sources of unobserved heterogeneity, thus outperforming the RPNB model. In addition, new insights into the interactions of safety factors (e.g., the joint safety effects of heavy trucks and pavement rutting depth) were obtained from the CRPNBHM model and these are expected to be beneficial in developing effective safety countermeasures. Results from this study demonstrated the CRPNBHM model to be a good alternative for crash frequency analysis, particularly when unobserved heterogeneity was detected.

In this paper, the effects of climbing lane on traffic safety were investigated using propensity scores and potential outcomes. Firstly, a binary logit model was developed to estimate propensity scores, and nearest neighbor matching technique with caliper was selected as the suitable matching method. Then, a random effects negative binominal (RENB) model was estimated to further explore the mechanism of safety benefits of climbing lanes. The results indicated that installing a climbing lane on steep uphill freeway segment can reduce the crash frequency, crashes per kilometer and crashes per 100 million vehicle-kilometers traveled by 18.27%, 19.49% and 17.41%, respectively. The safety effects of traffic volume, longitudinal grade and curvature did not change significantly for a segment before and after installing a climbing lane. However, the adverse effects of heavy truck proportion and distance along composite ascending grade on safety could be considerably mitigated by installing a climbing lane.

The study presented in this paper thoroughly investigated factors influencing driver injury severity in freeway single-vehicle crashes. Crash data from 2013 to 2017 for freeways in Heilongjiang Province, China was used. Elements of driver characteristics, environmental factors, roadway attributes and crash characteristics were considered. A heterogeneity-in-means mixed logit model was developed as an alternative to the frequently used multinomial logit model and mixed logit model to fully account for unobserved heterogeneity, particularly the heterogeneity resulting from driver characteristics. Results indicated that the mixed logit model with heterogeneity-in-means can provide a superior goodness-of-fit and offer more insights into factors of driver injury severities. By allowing means of random parameters in mixed logit model to be estimated functions of driver characteristics, a more general model structure for deeply tracking unobserved heterogeneity was constructed, and thereby the interactive effects between driver characteristics and other factors on driver injury severity were uncovered, such as: (1) female and senior drivers, darkness without lighting, collision with barriers or piers, lane-changing or merging maneuvers tend to increase the injury severity of drivers; (2) an experienced driver was associated with low probability of severe injuries; (3) low visibility could reduce injury severity, especially for experienced drivers; (4) a concrete barrier could aggravate the injury severity for senior drivers in particular. This study provided an insightful knowledge of mechanism of driver injury severity in single-vehicle crashes, and should be beneficial to develop corresponding effective countermeasures for protect drivers from being severely injured.

With the rapid increase in car ownership, urban transport systems are challenged by the overwhelming traffic demand and congestion. Dynamic prediction of traffic flows is of considerable significance for congestion mitigation and demand management. Real-time and precise prediction models are capable of analyzing traffic flow characteristics, predicting traffic flow trends, and motivating reasonable inductive actions. Considering the periodicity and variability of traffic flow and limitations of single prediction models, an adaptive hybrid model for predicting short-term traffic flow was proposed in this study. Firstly, the linear Autoregressive Integrated Moving Average (ARIMA) method and non-linear Wavelet Neural Network (WNN) method were used to predict traffic flow. Then, outputs of the two individual models were analyzed and combined by fuzzy logic and the weighted result was regarded as the final predicted traffic volume of the hybrid model. The results indicate that the hybrid model can offer better performance in predicting short-term traffic flow than the two single models either in stable or in fluctuating conditions. The relative error is within ±10%, showing that the proposed hybrid model is both accurate and reliable.

Signalized intersection has great roles in urban traffic system. The signal infrastructure and the driving behavior near the intersection are paramount factors that have significant impacts on traffic flow and energy consumption. In this paper, a speed guidance strategy is introduced into a car-following model to study the driving behavior and the fuel consumption in a single-lane road with multiple signalized intersections. The numerical results indicate that the proposed model can reduce the fuel consumption and the average stop times. The findings provide insightful guidance for the eco-driving strategies near the signalized intersections.

The freeway system of China is the longest in the world and planned to expand in the following years. However, the safety conditions of freeways are drawing increasing concerns both from the authorities and the public in the country. The safety design and management for freeways are urgently needed considering the current safety situation and great demands of new freeways. The study presented in this paper thoroughly investigated factors affecting safety using detailed data of crashes, traffic characteristics and freeway geometry. A random effects negative binomial (RENB) model was applied to account for spatial variations within groups together with a negative binomial (NB) model. The results indicated a better goodness of fit of RENB model than the NB model. In addition, a good number of factors significantly contributed to crash, such as truck proportion, presence of climbing lane, median barrier offset, curvature and longitudinal grade, were identified. This study was expected to provide a better understanding of how traffic condition and freeway design affect safety and should be useful to freeway engineers to design safe freeways, or develop effective safety countermeasures.