Accid Anal Prev [JOURNAL]
Sun
200 papers
A safety field-based surrogate measure for quantifying bikelane conflict risk considering physical and behavioral dynamics.
Liang X, Ding J, Zheng L
Accid Anal Prev · 2026 Sep · PMID 42314363 · Publisher ↗
While offering separation between vulnerable road users and motorized vehicles, the presence of non-motorized traffic facilities, such as bikelanes, introduces complex interaction dynamics among pedestrians, conventional...
While offering separation between vulnerable road users and motorized vehicles, the presence of non-motorized traffic facilities, such as bikelanes, introduces complex interaction dynamics among pedestrians, conventional bicycles, and e-bikes. Existing surrogate measures inadequately quantify conflict risks in these environments due to oversimplified assumptions and limited consideration of contextual factors. This study proposes a novel safety field-based surrogate measure (Bikelane Safety Field-based Measure, BSFM) that integrates physical dynamics (i.e., road environmental characteristics, kinematic interactions) and behavioral dynamics (i.e., psychological comfort, risk perception, evasion behaviors) to quantify bikelane conflict risk. Utilizing drone-collected trajectory data (203 conflict groups, 37,652 s) from three Tianjin intersections, a bikelane-specific safety field model was developed. Subsequently, the BSFM was proposed, and the threshold of the surrogate measure was determined using extreme value theory. Validation of the BSFM yielded the following key findings: (i) The BSFM demonstrated superior conflict identification recall (81.3%) compared to Time-to-Collision (TTC) (34.5%) and Projected Time-to-Collision (PTTC) (41.4%). (ii) Significant sensitivity to evasive actions was observed, with Kolmogorov-Smirnov and Mann-Whitney U tests confirming statistically significant changes in BSFM values during swerving and deceleration maneuvers (p < 0.001). (iii) Real-time risk tracking was effectively achieved through dynamic visualizations of the safety performance envelope. (iv) The model exhibited robust applicability across diverse conflict participants, including pedestrians, bicycles, and e-bikes. The BSFM provides a validated framework for real-time safety assessment in shared micro-mobility environments, advancing proactive traffic management strategies.
Enhancing safety in freeway on-ramp merging: Takeover warning through dynamic trust and risk perception modelling considering driving style.
Yao M, Wang J, Gong J … +3 more , Niu Y, Liu M, Yin G
Accid Anal Prev · 2026 Sep · PMID 42314362 · Publisher ↗
Freeway on-ramp merging represents a high risk conflict zone in which control transition failures in conditionally automated driving (AD) can contribute to severe collisions. Such events are often linked to a mismatch be...
Freeway on-ramp merging represents a high risk conflict zone in which control transition failures in conditionally automated driving (AD) can contribute to severe collisions. Such events are often linked to a mismatch between rapidly shrinking safety margins and driver cognitive readiness at takeover. Conventional takeover warning strategies that rely on static kinematic thresholds may not account for the driver dynamic pre-crash state including trust, risk perception, and driving style. This limitation can increase the likelihood of missed alarms and nuisance alarms and can reduce the effectiveness of warnings for accident prevention. To address this safety gap, this study proposes a personalized accident prevention framework that identifies high risk states and triggers predictive intervention before physical conflict becomes critical. The method integrates a Kalman filter based dynamic trust estimator with a learned risk perception module to construct a driving style sensitive joint trigger mechanism. The mechanism issues warnings when the driver trust state deviates from a personalized trust interval that represents a safe operating envelope under prevailing risk. Two independent simulator experiments were conducted to calibrate style specific trust intervals and to validate the proposed strategy in complex merging scenarios. Comparative results show improved safety outcomes relative to a fixed threshold baseline. The proposed strategy reduced mean collision rate from 40.46% to 17.05% and increased correct takeover rate from 51.23% to 64.22%. These findings indicate that integrating dynamic human state estimation and driving style heterogeneity into takeover warning logic can support prevention of takeover related accidents and can enhance safety resilience in automated merging operations.
Complex but solvable: towards a cognitive human-like risk-identification model for AV-HV mixed traffic.
Xie J, Li J, Zhu Y … +2 more , Wu S, Xia Y
Accid Anal Prev · 2026 Sep · PMID 42308565 · Publisher ↗
Human-driven vehicles (HVs) in mixed traffic environments exhibit significant behavioral uncertainty and heterogeneous driving patterns, posing substantial challenges for autonomous vehicles (AVs) in risk identification....
Human-driven vehicles (HVs) in mixed traffic environments exhibit significant behavioral uncertainty and heterogeneous driving patterns, posing substantial challenges for autonomous vehicles (AVs) in risk identification. This challenge is particularly pronounced in interweaving areas with high traffic dynamics, dense interactions, and complex operational conditions. Existing risk identification methods mainly rely on predefined scenarios or static decision rules, which often lead to limited generalization capability, slow response to extreme risks, and insufficient interpretability in complex mixed traffic environments. To address these limitations, this study draws inspiration from the adaptive learning and consultation behaviors of humans and constructs a cognitive optimization method based on the human behavior-based optimization (HBBO) algorithm. Based on this, a cognitive human-like risk-identification model (CHRIM) is proposed for autonomous vehicle-human-driven vehicle (AV-HV) interactions in interweaving areas. The model simulates key cognitive abilities of human drivers, including abstract understanding of driving features, retrospective reasoning of risk levels, and cognitive optimization of risk identification strategies, enabling adaptive evolution and dynamic updating of risk identification mechanisms. Experiments based on real trajectory data collected from urban expressway interweaving area in Chongqing demonstrate that the proposed method achieves superior performance in risk identification accuracy and robustness. The accuracy (ACC), matthews correlation coefficient (MCC), and kappa (KAP) reach 0.9543, 0.8906, and 0.8782, respectively, significantly outperforming several mainstream methods, including the fine tree model (FTM), random subspace model (RSM), efficient logistic regression model (ELRM), and neural networks (NN). In addition, an explainable analysis framework is developed to provide intuitive insights into human-like risk identification of AVs and to reveal the risk evolution process in interweaving area.
A hybrid latent class analysis and association rule mining framework to identify injury-related risk patterns in level 2 advanced driver assistance system crashes.
Wang J, Harper CD, Hendrickson C
Accid Anal Prev · 2026 Sep · PMID 42296798 · Publisher ↗
Automated driving features could improve road safety by reducing the number of crashes caused by human error. However, as the market penetration of Level 2 advanced driver assistance systems (ADAS) increases, so does the...
Automated driving features could improve road safety by reducing the number of crashes caused by human error. However, as the market penetration of Level 2 advanced driver assistance systems (ADAS) increases, so does the number of crashes involving these technologies. This study proposes a data-driven framework combining latent class analysis, association rule mining, and logistic regression to analyze 617 Level 2 ADAS crashes reported to the National Highway Traffic Safety Administration between 2019 and 2026. Two latent classes are identified, which are lower-speed local road crashes and high-speed highway crashes. Association rule mining and logistic regression are conducted within each latent class, revealing substantial heterogeneity in injury-related crash patterns. In lower-speed local road crashes, injury occurrences are mainly associated with frontal impacts, intersection-related crashes, and moderate-speed configurations. In high-speed highway crashes, injury occurrences are mainly associated with high-speed configurations, fixed objects, light-to-medium duty vehicles, frontal impacts, and morning or late-night conditions. Practical applications include placing greater emphasis in scenario-based testing for combinations involving moderate and high pre-crash speeds, frontal contact, fixed-object crashes, light-to-medium duty vehicle crash partners, and late-night conditions. These findings may help manufacturers refine speed-aware system-use guidance, driver warnings, and operational design constraints for Level 2 ADAS-equipped vehicles. Overall, this study demonstrates the value of class-specific analysis for uncovering heterogeneous injury-associated patterns in Level 2 ADAS crashes.
Examining weekday-weekend variations in factors affecting pedestrian crashes: A geospatial explainable machine learning framework.
Wang Z, Fan W
Accid Anal Prev · 2026 Sep · PMID 42296797 · Publisher ↗
Temporal shifts in travel demand and activity patterns between weekdays and weekends substantially alter pedestrian exposure and crash occurrence mechanisms, implying that the contributing factors differ accordingly. In...
Temporal shifts in travel demand and activity patterns between weekdays and weekends substantially alter pedestrian exposure and crash occurrence mechanisms, implying that the contributing factors differ accordingly. In addition, although nonlinear effects and spatial heterogeneity have been widely investigated, context-dependent interactions among multiple factors in pedestrian crashes remain insufficiently understood. Ignoring such interdependencies may obscure the influence of certain factors under specific conditions and challenge the effectiveness of policies. Therefore, this study investigates the weekday-weekend variations in factors affecting pedestrian crash density by jointly accounting for nonlinear threshold effects, context-dependent effects, and spatial heterogeneity. Four years of pedestrian-vehicle crash data (2021-2024) from Mecklenburg County, North Carolina, were collected and aggregated at the census tract level, with crashes stratified into weekday and weekend periods. Then, a geospatial explainable machine learning framework integrating an XGBoost-Tweedie model tailored for zero-inflated crash data with two explainable artificial intelligence approaches, SHAP and GeoShapley, was applied to uncover these complex effects. Model comparison results indicate that the XGBoost-Tweedie model outperforms both the traditional and geographical versions of Random Forest, XGBoost, and LightGBM. Results further reveal that infrastructure characteristics play a dominant role in pedestrian crashes during both weekdays and weekends. However, spatial inequities reflected by sociodemographic characteristics (e.g., Black ratio and poverty ratio) exert stronger influences on weekend crashes. Moreover, the effects of these factors exhibit pronounced variations across different thresholds, spatial locations, and contextual settings. These findings provide critical insights for developing period-specific, location-specific, and context-specific countermeasures to enhance pedestrian safety.
Nonlinear coupling of cognitive distraction and potential vehicle-pedestrian conflicts: Dynamic quantification of distraction level and driving risk.
Peng J, Zhan B, Yuan H … +2 more , Ren C, Lei Z
Accid Anal Prev · 2026 Sep · PMID 42284979 · Publisher ↗
Cognitive distraction and potential vehicle-pedestrian conflicts are common crash risk factors, and their co-occurrence may compound the risk. Prior research has not fully characterized behavior changes, risk evolution o...
Cognitive distraction and potential vehicle-pedestrian conflicts are common crash risk factors, and their co-occurrence may compound the risk. Prior research has not fully characterized behavior changes, risk evolution over time, or interaction mechanisms in this combined setting. To address these gaps, we conducted a driving-simulator study with 38 licensed drivers, examining their behavioral responses to potential vehicle-pedestrian conflict scenarios under varying levels of cognitive distraction. We employed a sliding time-window approach combined with machine-learning methods to dynamically identify driving states. Multi-source indicators, including visual perception, vehicle kinematics, and response decision behavior, were integrated to develop a real-time quantitative model linking distraction level to driving risk. Results indicated that under baseline driving, drivers exhibited higher visual workload, broader environmental scanning, and more conservative control in the covert occlusion-zone scenario than in the overt noncompliant-crossing scenario. As secondary-task complexity increased, gaze became progressively more concentrated in the near-forward field, and fixation and braking response times lengthened, and control became more disordered. Consequently, behavioral differences between scenarios narrowed, and coping strategies converged under high-complexity tasks. Real-time estimates showed stepwise increases in both distraction level and driving risk as task complexity rose. Under high-complexity conditions, risk became more volatile, and peak-risk differences between scenarios nearly disappeared. These findings clarify the dynamic coupling between cognitive distraction and potential vehicle-pedestrian conflict risk, informing driver-state-aware assistance systems and context-adaptive intervention strategies.
Merging behavior under varying work zone sign scenarios: A heterogeneity-based analysis.
Yan X, Xie Y, Bhuyan Z … +3 more , Xiang B, Wu G, Shirazi M
Accid Anal Prev · 2026 Sep · PMID 42251814 · Publisher ↗
Proper merging behavior is critical for work zone safety. A thorough understanding of the factors influencing how vehicles merge under different sign scenarios is essential for developing and evaluating effective safety...
Proper merging behavior is critical for work zone safety. A thorough understanding of the factors influencing how vehicles merge under different sign scenarios is essential for developing and evaluating effective safety strategies. This study investigates merging behaviors at highway work zones under different flashing speed limit sign (FSLS) and portable changeable message sign (PCMS) scenarios. Two groups of random parameters logit models with heterogeneity in means and variances are estimated for different FSLS and PCMS configurations. Using real-world radar, thermal camera, and meteorological data, three categories of merging behavior are defined as outcome variables: risky merge, somewhat risky merge, and safe merge. Multiple traffic and environmental characteristics are included as explanatory variables. Likelihood ratio tests indicate the determinants of merging behaviors vary across traffic control scenarios, which is further validated through comparisons between out-of-sample and within-sample predictions. The results reveal that the determinants of merging behaviors vary across sign scenarios and that ignoring scenario interactions may introduce systematic prediction bias. These findings highlight the importance of accounting for scenario-specific variations when identifying precursors and forecasting merging behaviors. The results also suggest that PCMS message design should consider potential interactions with raised or flashing FSLS, as simultaneous visual stimuli may increase drivers' cognitive load. In addition, adapting traffic control strategies to environmental conditions, particularly during adverse weather or nighttime scenarios, may help promote more cautious merging behavior. Finally, given the scenario-dependent variations observed in model parameters, further research is needed to examine how the effectiveness of FSLS and PCMS evolves over time. Future studies could also explore spatiotemporal stability using larger datasets across multiple work zones.
Human-automation interaction shapes safety performance across automation levels in safety-critical scenarios.
Qin D, Shen J, He Z … +6 more , Li Q, Di Q, Zhang X, Zhan Z, Yu H, Nie B
Accid Anal Prev · 2026 Sep · PMID 42247825 · Publisher ↗
Automated driving is widely expected to improve road safety, yet its realized safety performance in safety-critical scenarios remains uncertain because outcomes depend not only on system capability, but also on how drive...
Automated driving is widely expected to improve road safety, yet its realized safety performance in safety-critical scenarios remains uncertain because outcomes depend not only on system capability, but also on how drivers perceive risk, intervene, and interact with automation. This study establishes a unified driver-in-the-loop experimental framework for evaluating safety performance under human-automation interaction across multiple automation levels. Using high-fidelity driving simulator experiments covering representative safety-critical scenarios and SAE Levels 2-4, we collected a large-scale dataset of driver-automation interactions and safety outcomes. The results show that realized safety in safety-critical events is jointly shaped by automation capability and driver intervention behavior. As automation capability increased, driver intervention became less frequent and generally later, while overall safety performance improved. At the same time, a different cross-level regularity emerged among cases involving driver intervention: under the present experimental conditions, the collision risk conditional on intervention remained at a relatively stable non-zero level (approximately 26%), and both the intervention-onset risk state and the controllability boundary showed broadly similar patterns across levels. These findings indicate that automation level mainly changes whether and when drivers intervene, while the risk state at intervention onset and the overall effectiveness pattern of intervention remain broadly stable across levels. Model-based validation further showed that collision risk is primarily associated with the driver's risk state at intervention onset and response latency. Overall, this study identifies a cross-level regularity in driver-automation interaction during safety-critical events and provides a driver-centered basis for understanding takeover limits and intervention-conditioned collision risk, with implications for the human-centered evaluation and design of safer automated driving systems.
TRIDENT: A multi-task, triple-branch deep learning framework for EEG-based recognition, severity estimation, and future high-anger prediction in an on-road Wizard-of-Oz paradigm.
Guo Y, Gao Z, Zhang T … +3 more , Deng L, Cai J, Xiao H
Accid Anal Prev · 2026 Sep · PMID 42235288 · Publisher ↗
Driving anger is strongly associated with aggressive driving and elevated crash risk. However, continuous modeling of graded anger-related affective states from physiological signals under controlled on-road exposure rem...
Driving anger is strongly associated with aggressive driving and elevated crash risk. However, continuous modeling of graded anger-related affective states from physiological signals under controlled on-road exposure remains underexplored. In this study, we investigated driving-anger-related affective states elicited in an on-road Wizard-of-Oz (WoZ) paradigm, in which 24 licensed participants sat in the front passenger seat, imagined themselves as the driver, and were exposed to scripted traffic-conflict maneuvers executed by a trained safety driver. Using synchronized 32-channel electroencephalography (EEG), self-reports, and contextual information, we constructed a four-level anger-state dataset. Building on this dataset, we introduce TRIDENT, a multi-task deep learning model designed to simultaneously recognize four levels of anger states, estimate continuous anger severity (0-100), and predict future high-anger states. TRIDENT integrates multi-scale temporal convolutions, brain-network representations, and sequence modeling to capture complementary spatiotemporal patterns of anger dynamics. Experimental results show that TRIDENT significantly outperforms representative baseline EEG emotion recognition models, achieving up to 85% accuracy in four-class anger-state classification and 87% accuracy in predicting future high-anger states. Scalp topographies and cortical source localization analyses further reveal anger-level-dependent changes in prefrontal, temporal, and limbic brain networks. These findings provide a physiologically grounded perspective on anger-related neural dynamics under controlled on-road conflict exposure and have implications for emotion-aware in-vehicle interfaces and personalized intervention design. Code will be made available upon acceptance at: https://github.com/tianyaz719/TRIDENT.
MRISce: an interactive autonomous driving test scenario generation method based on multi-agent reinforcement learning.
Li J, Wang R, Zhu Y … +2 more , Zhang M, Zhao X
Accid Anal Prev · 2026 Sep · PMID 42235287 · Publisher ↗
Road safety remains the primary objective in the development and validation of autonomous driving systems. Scenario-based simulation testing provides an efficient and controllable method for safety verification. However,...
Road safety remains the primary objective in the development and validation of autonomous driving systems. Scenario-based simulation testing provides an efficient and controllable method for safety verification. However, the lack of dynamic interaction between the background vehicle and the vehicle under test in the current autonomous driving simulation test scenarios results in insufficient interactivity. To address this issue, this study proposes MRISce, a dynamic interactive test scenario generation method based on multi-agent reinforcement learning, which allows for the simulation of more complex dynamic traffic scenarios by developing an interactive driving strategy for the background vehicle, resulting in more realistic interactive driving behavior. Initially, a vision-based dynamic driving model is created. The upgraded Level-K multi-agent reinforcement learning framework is then used to generate three distinct types of background vehicle driving strategies with varying interaction degrees. Finally, a closed-loop simulation platform is developed, in which the background vehicle driving model is used to generate dynamic interactive driving scenarios, with a typical intersection test scenario serving as an example to verify MRISce experimentally. The experimental results show that the MRISce-generated test scenarios when compared to the test scenarios constructed under the traditional scheme, bring a maximum of 27.6% increase in the collision rate of the vehicle under test, approximately 61.4% of the delay in reaching the destination, 48% of the decrease in the arrival rate, a 60.1% advance in the time of the first collision, and an improvement of nearly 8 times in the interactivity indices. The findings suggest that MRISce may greatly improve the interaction characteristics of test scenarios while effectively evaluating the vehicle's performance in increasingly demanding dynamic conditions.
Investigating heterogeneous human factor effects on crash severity: A counterfactual study in mountainous freeways.
Li J, Guo F, Yang W … +2 more , Guo Y, Chen Y
Accid Anal Prev · 2026 Sep · PMID 42229309 · Publisher ↗
Accurately estimating the causal effect of human factors on crash severity in mountainous freeways is pivotal for developing effective safety strategies. Although previous studies have investigated human factors, they ty...
Accurately estimating the causal effect of human factors on crash severity in mountainous freeways is pivotal for developing effective safety strategies. Although previous studies have investigated human factors, they typically focus on estimating average effects under specific conditions, often conflating statistical correlations with causal relationships. Consequently, the underlying causal mechanisms in freeway crashes remain unclear. To address this limitation, this study proposes a dynamic weighted double machine learning framework that integrates LightGBM and XGBoost models to estimate confounder-outcome and treatment-outcome relationships. By optimizing model weights and using non-human-caused crashes as a control group, the effects of five human factors are isolated through counterfactual reasoning. The heterogeneous treatment effects of human factors on crash severity are quantified, and causal relationships are analyzed across variables such as weather, slope, truck traffic volume, and vehicle type. The results reveal significant heterogeneity in crash severity attributable to human factors compared to non-human factors. Inadequate safety distance often co-occurs with other high-risk conditions, amplifying crash severity. Reversing behavior is particularly sensitive to weather conditions. The causal pathways of distracted or fatigued driving and driving in the wrong lane are influenced by daily truck traffic volume and road slope. Additionally, interactions between slope and vehicle type significantly affect the severity of overloaded crashes. These findings underscore the need for targeted interventions addressing specific human factors in high-risk scenarios. Consequently, enforcement against reversing and overloading on steep slopes with high truck volume should be intensified, and heavy trucks should be restricted to right lanes on high-risk segments.
Scenario and leading vehicle-dependent car-following behaviors and risks: a real-world comparison of automated and human-driven vehicles.
Sun L, Zhao X, Li P
Accid Anal Prev · 2026 Sep · PMID 42229308 · Publisher ↗
The increasing deployment of automated vehicles (AVs) has raised safety concerns, because interactions between AVs and human-driven vehicles (HVs) may introduce new longitudinal risk patterns in mixed traffic. Using real...
The increasing deployment of automated vehicles (AVs) has raised safety concerns, because interactions between AVs and human-driven vehicles (HVs) may introduce new longitudinal risk patterns in mixed traffic. Using real-world data from the OpenACC database, this study empirically examines how scenarios and leading vehicle types influence car-following behavior and associated risks. Descriptive statistics reveal that HVs are highly sensitive to both factors: they exhibit bimodal speed distributions on highways, maintain longer headways in test scenarios, show heightened vigilance when following AVs, and display more relaxed behavior-including occasional speeding-when following other HVs. In contrast, AVs demonstrate consistent, algorithm-driven responses to the dynamics of the leading vehicle. Mixed-effects models confirm that scenario and leading vehicle type have stronger effects on behavioral and risk indicators for HVs than for AVs. Although AVs exhibit a higher risk rate, reflected in more frequent short headways, they maintain lower risk severity and adhere more closely to low Time-to-Collision (TTC) thresholds, suggesting proactive risk management. For HVs, risk is significantly higher in highway scenarios and when following an AV. The highest risk occurs when an AV follows an HV, highlighting the safety challenge posed by the unpredictability of human drivers. These findings underscore that interactions among automation level, scenario, and car-following pair types are crucial for accurately assessing and mitigating collision risks in mixed traffic environments.
Real-time lane-level abnormal traffic detection on freeways using sparse telematics data.
Liang S, Ma C, Li P … +8 more , Shi H, Liu J, Zhou H, Long K, Cao B, Szymkowski T, Shi X, Li X
Accid Anal Prev · 2026 Sep · PMID 42224903 · Publisher ↗
Real-time abnormal traffic detection is critical in intelligent transportation systems because traditional abnormal traffic notifications often suffer delays and lack specific, lane-level location information, which can...
Real-time abnormal traffic detection is critical in intelligent transportation systems because traditional abnormal traffic notifications often suffer delays and lack specific, lane-level location information, which can lead to safety risks and economic losses. This paper proposes a real-time, lane-level abnormal traffic detection approach for freeways that only leverages sparse telematics trajectory data. In the offline stage, the historical trajectories are discretized into spatial cells using vector cross-product techniques, and then are used to estimate a vehicle intention distribution and select an alert threshold by maximizing the F1-score against official crash reports used as proxy labels for abnormal events. In the online stage, incoming real-time telematics records are mapped to these cells and scored for three modules: transition anomalies, speed deviations, and lateral maneuver risks, with scores accumulated into a cell-specific risk map. When any cell's risk exceeds the alert threshold, the system issues a prompt warning. Relying solely on telematics data, this real-time and low-cost solution is empirically evaluated using these report-based labels in Wisconsin, achieving a 75% identification rate with accurate lane-level localization, an overall accuracy of 96%, an F1-score of 0.84, and a false alarm rate of only 0.6% among non-crash cases, while also detecting 13% of crashes more than 3 min before the notification time.
Integrating pavement condition records with LLM-based crash narrative analysis for pavement safety assessment.
Gottimukkala SV, Gao L, Gharaibeh N … +4 more , Tarnini Y, Talebzadeh M, Kutela B, Katale TS
Accid Anal Prev · 2026 Sep · PMID 42217472 · Publisher ↗
This paper proposes an integrated framework that couples Large Language Models (LLMs)-based crash narrative analysis with quantile regression to identify and quantify pavement-related crash risk. The LLM component conver...
This paper proposes an integrated framework that couples Large Language Models (LLMs)-based crash narrative analysis with quantile regression to identify and quantify pavement-related crash risk. The LLM component converts unstructured police narratives into structured, mechanism-specific labels (e.g., hydroplaning, curve-related loss of control), which enables outcomes that are directly linked to pavement and roadway-surface conditions that are often missing from conventional structured crash fields. These LLM-derived mechanism labels are then matched to segment-level pavement condition, friction, texture, traffic exposure, and geometric characteristics and modeled using quantile regression to characterize how covariate effects vary across the full distribution of crash risk rather than only at the mean. A case study using over 24,000 police crash narratives linked to a pavement management dataset of approximately 180,000 data records demonstrates strong associations between friction/texture measures and wet-pavement crash mechanisms. These results can help transportation agencies select candidate pavement-safety projects by identifying pavement conditions associated with elevated crash risk and prioritizing targeted, cost-effective countermeasures.
Association between traffic reoffending and type of official points recovery course (partial, full, and court-mandated): Evidence from a three-year follow-up.
Castro C, Padilla FM, Pacheco-Unguetti AP
Accid Anal Prev · 2026 Sep · PMID 42202513 · Publisher ↗
Traffic recidivism is a major contributor to road accidents, injuries, and fatalities, highlighting the need for effective interventions aimed at reducing risky driving behaviour. This study analyses objective records of...
Traffic recidivism is a major contributor to road accidents, injuries, and fatalities, highlighting the need for effective interventions aimed at reducing risky driving behaviour. This study analyses objective records of traffic sanctions and penalty point losses to examine the long-term effectiveness (three-year follow-up) of official driving licence points recovery courses within the Spanish Points Penalty System (PPS) in preventing repeat traffic offences. The sample comprised 140 offender drivers who completed one of three official recovery courses: 55 attended a partial recovery course, 35 completed a full recovery course after losing all licence points, and 50 participated in a court-mandated recovery course. Objective data on traffic sanctions and penalty point losses were collected for the three years following course completion. Descriptive analyses indicated high overall recidivism. Data provided by the Dirección General de Tráfico (DGT) showed that 71.43% of participants received at least one traffic sanction, and 47.86% lost at least one penalty point during the follow-up period. However, reoffending patterns varied significantly across course types. Drivers attending partial recovery courses showed the highest recidivism, with 83.64% receiving at least one sanction and 72.73% losing penalty points. Participants in full recovery courses also exhibited high sanction rates (80%), although a substantially lower proportion experienced further point loss (28.57%). In contrast, drivers enrolled in court-mandated courses showed lower recidivism rates, with 52% receiving at least one sanction and 34% experiencing point loss. Overall, long-term reoffending was more strongly associated with the type of course attended than with sociodemographic characteristics or substance use. Court-mandated courses were associated with lower levels of sanctions and penalty point losses over the three-year follow-up, potentially reflecting the stronger legal consequences associated with these interventions. Full recovery courses showed patterns consistent with partial behavioural adaptation, with participants tending to shift towards less severe infractions rather than achieving full compliance. Finally, partial recovery courses appeared less effective, possibly due to lower perceived deterrent value or their use as a procedural mechanism for point restoration rather than as a rehabilitative intervention. From this perspective, the higher effectiveness of court-mandated courses may be explained not only by their educational and punitive components, but also by the broader legal consequences experienced by offenders, including community service orders or custodial sentences.
An efficient methodology for modeling imbalanced traffic crashes through deep learning techniques.
Irandoost A, Ghaemi M, Shad R … +1 more , Ziaee SA
Accid Anal Prev · 2026 Sep · PMID 42190349 · Publisher ↗
Accurately predicting crash injury severity is crucial for enhancing traffic safety. However, crash datasets are often highly imbalanced because severe injuries are infrequent, which can bias predictive models. Addressin...
Accurately predicting crash injury severity is crucial for enhancing traffic safety. However, crash datasets are often highly imbalanced because severe injuries are infrequent, which can bias predictive models. Addressing these challenges, the present study develops a hybrid methodology named ENN-CTGAN. ENN-CTGAN is designed to address class imbalance and overcome the limitations of traditional resampling techniques. A hybrid LSTM-GRU model was developed to predict injury severity, and its performance was compared with LSTM, GRU, CNN, MLP, XGBoost and Random Forest models. The study investigates different synthetic data ratios (1:1, 1:2, 1:4, and 1:6) to identify the optimal ratio for each prediction model. It also proposes a framework that combines mutual information differences with model efficiency to assess the quality of the generated synthetic data. The performance of ENN-CTGAN was compared with other resampling techniques, including SMOTE, Random Oversampling, and Random Undersampling. Prediction performance was evaluated with AUC, G-mean, Sensitivity, Specificity, and Accuracy. Results demonstrate that the proposed two-stage framework outperforms conventional resampling approaches across multiple predictive models. Under fully balanced data (1:1 ratio) the hybrid LSTM-GRU model achieved the best performance within the ENN-CTGAN framework with a G-mean of 0.5452. Among all evaluated ratio configurations and resampling methods, the XGBoost classifier under the proposed ENN-CTGAN framework achieved the highest performance at a 1:4 synthetic proportion (G-mean = 0.5643). These findings highlight the methodological advantage of the ENN-CTGAN design and underscore the importance of empirical, data-driven ratio optimization rather than assuming a fixed balancing strategy in imbalanced crash severity modeling.
Modeling driver lane-changing aggressiveness under target-lane interference: A Bayesian approach using naturalistic trajectory data.
Shen R, Ma L, Yan X … +1 more , Yang Z
Accid Anal Prev · 2026 Sep · PMID 42177847 · Publisher ↗
Aggressive lane-changing behavior on highways can induce sharp lateral movements and significant speed variations, posing considerable traffic safety risks. Maximum lateral velocity effectively reflects both the vehicle'...
Aggressive lane-changing behavior on highways can induce sharp lateral movements and significant speed variations, posing considerable traffic safety risks. Maximum lateral velocity effectively reflects both the vehicle's lateral motion and the intensity of the driver's lane-change maneuver, making it a key metric for analyzing lane-changing behavior. In this study, 1,646 lane-changing events were extracted from the real-world vehicle trajectory collected by CQSkyEyes, and their basic characteristics were analyzed using descriptive statistics. The Bayesian model was developed to examine how lane-change intensity varies under different environmental and driving conditions. The results indicate that under complex weather conditions, drivers tend to adopt more conservative lane-changing strategies, reflected in reduced maximum lateral velocity, especially when facing hazardous time-to-collision (TTC) levels. Moreover, lane change behavior is shaped by surrounding-vehicle interaction metrics; notably, the approach of a vehicle in the target lane increases the likelihood of aggressive maneuvers. These findings highlight maximum lateral velocity as a robust quantitative indicator of driver behavior, offering actionable implications for traffic safety management and autonomous driving system design.
A full Bayesian random parameters Negative Binomial-Lindley model for fatal pedestrian crash frequency on rural highways.
Aman P, Tiwari G, Rao KR
Accid Anal Prev · 2026 Sep · PMID 42176507 · Publisher ↗
Pedestrian safety on rural highways remains a pressing concern in low- and middle-income countries (LMIC), where highways frequently pass through settlements with substantial pedestrian activity but limited dedicated ped...
Pedestrian safety on rural highways remains a pressing concern in low- and middle-income countries (LMIC), where highways frequently pass through settlements with substantial pedestrian activity but limited dedicated pedestrian infrastructures and prevalence of high-speed motorized traffic. This study investigates fatal pedestrian crash frequency on such roads using a fine-resolution segment-level approach. Five-year fatal pedestrian crash data (2017-2022, excluding 2020), comprising 337 pedestrian-involved crashes across six rural highway corridors, were analyzed by dividing the entire road network into 2,881 equal-length segments. Crash data were integrated with detailed segment-level information on traffic exposure, vehicular speeds, roadway geometry, roadside environment, landuse, and population exposure. Four count data models were estimated within a Full Bayesian framework to account for overdispersion, excess zeros, and unobserved heterogeneity, and the best-fitting model was identified for pedestrian crash frequency. Model performance was evaluated using the Deviance Information Criterion, predictive accuracies, and cumulative residual plots, which support the superiority of the Random Parameter Negative Binomial-Lindley (RPNB-L) model over its counterparts. Results indicate that the presence of junctions, settlements, landuse, flyover transition zones, canals/bridges/culverts, median gaps, service roads, minor accesses, and pedestrian activity generators such as bus stops, schools, fuel stations, and roadside eateries primarily drives pedestrian crash risk on rural highways. Population exposure emerges as a robust predictor of risk, while observed pedestrian volumes exhibit a safety-in-numbers effect. The findings emphasize the need for segment-level pedestrian safety strategies on high-speed rural highways and provide empirical evidence to support targeted infrastructure design, access management, and roadside environment interventions in LMICs.
Research on formation vehicles cooperative strategy based on reinforcement learning and GANs at uncontrolled intersections.
Tong W, Chaochun Y, Yingfeng C … +4 more , Youguo H, Jie S, Xinkai W, Shuofeng W
Accid Anal Prev · 2026 Sep · PMID 42172805 · Publisher ↗
This paper presents an innovative model of safe driving for formation vehicles while passing potential traffic accident areas and a special strategy to optimize the velocity and acceleration based on Reinforcement Learni...
This paper presents an innovative model of safe driving for formation vehicles while passing potential traffic accident areas and a special strategy to optimize the velocity and acceleration based on Reinforcement Learning (RL) and Generative Adversarial Networks (GANs). The concept of occluded scenes and uncontrolled intersections are described and the parameters are detailed. The evasion mechanism underlying risk-aware driving is systematically analyzed. Based on this analysis, a safety-oriented driving strategy for occluded scenarios is proposed to generate preliminary yet reliable reference velocity and acceleration for formation vehicles. Reinforcement learning model for uncontrolled intersection is trained and tested to optimize the formation driving strategy. Furthermore, generative adversarial networks is used to enrich driving scenarios and enhance the safety and efficiency of strategies presented. Experimental results in simulation show that the proposed formation vehicles driving strategy can improve the stability of formation when facing uncontrolled intersections.
How should drivers' use of Automated Lane Keeping Systems (ALKS) be assessed? A study with experienced driving assessors in a Wizard-of-Oz vehicle.
Bazilinskyy P, Heikoop DD, Verstegen R … +2 more , Martens MH, de Winter JCF
Accid Anal Prev · 2026 Sep · PMID 42172804 · Publisher ↗
This study aims to contribute to guidelines for driver licensing organizations on assessing driver competence in using Level 3 Automated Lane Keeping Systems (ALKS), based on an on-road experiment with eight professional...
This study aims to contribute to guidelines for driver licensing organizations on assessing driver competence in using Level 3 Automated Lane Keeping Systems (ALKS), based on an on-road experiment with eight professional driving assessors (i.e., expert driving examiners who train examiner candidates; 6 males, 2 females, all driving more than 20,000 km per year) in a Wizard-of-Oz vehicle. Using a think-aloud protocol, we captured cognitive processes during system supervision and take-over requests (TORs) in real-world traffic jams. A large language model (LLM)-based thematic analysis of transcripts revealed five themes: (1) Requirement for immediate environmental assessment, (2) Requirement for causal understanding, (3) Requirement for proactive intervention to maintain traffic flow, (4) Requirement for continuous "supervisor" engagement, and (5) Physical ergonomics and mode awareness. These findings indicate that, at least during short-duration usage, drivers do not simply rely on the system to disengage from driving; instead, they maintain active monitoring, physical readiness, and anticipatory skills. These observations blur the distinction between Level 2 and Level 3 automation, as the expert participants in this study generally remained attentive rather than adopting the 'mind-off' state that Level 3 theoretically allows. In conclusion, assessing ALKS usage involves not only evaluating a driver's reaction to a TOR but also judging their performance as a systems manager responsible for anticipating conflicts and smoothly executing control transitions.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe