Accid Anal Prev [JOURNAL]
Sun
200 papers
Modeling road-segment-level speeding risk of new energy vehicle taxis using a multistage framework with spatial spillover, endogeneity, and nonlinear effects.
Qu L, Jiang X, Fu C … +2 more , Yu Q, Zhou Y
Accid Anal Prev · 2026 Jul · PMID 42391868 · Publisher ↗
Speeding poses a critical risk to urban traffic safety, especially for new energy vehicle taxis, whose capacity of high torque and rapid acceleration may exacerbate this risk. To investigate the speeding risk of new ener...
Speeding poses a critical risk to urban traffic safety, especially for new energy vehicle taxis, whose capacity of high torque and rapid acceleration may exacerbate this risk. To investigate the speeding risk of new energy vehicle taxis, this study proposes a multistage analytical framework, which combines endogeneity-adjustment methods and machine learning to account for spatial spillovers, endogeneity, and nonlinear effects. The framework employs geographically weighted generalized propensity score matching to capture nearby road segments' spillover effects and balance observable endogeneity, and combines spatial instrumental variables with two-stage residual inclusion method to account for potential unobservable endogeneity. Besides, the study develops a generalized speeding risk index that incorporates speeding frequency, severity, and exposure to comprehensively measure the speeding risk at the road segment level. Further, three Extreme Gradient Boosting models with Shapley Additive Explanations values and Accumulated Local Effects plots are developed to capture the nonlinear relationships between speeding risk and four categories of influencing factors (i.e., traffic conditions, road characteristics, traffic management strategies, and points-of-interest density) and enhance model interpretability. The proposed framework is applied using GPS trajectory data of new energy vehicle taxis in Chengdu, China. Results show that the proposed framework significantly outperforms conventional geographically weighted regression and linear regression models. Medium traffic flow, moderate intersection spacing, 4-5-lane segments, low commercial points-of-interest density, and roads near charging stations are associated with higher speeding risk, whereas no-parking rules, higher residential points-of-interest density, physical separators, and higher speed-limit categories are associated with lower risk. The findings further indicate that traffic management strategies become more influential after accounting for spatial spillover effects and correcting endogeneity, highlighting the importance of structurally aware modeling for urban traffic safety analysis. The study provides theoretical insights and practical implications for urban traffic safety management in the era of vehicle electrification.
Role of streetscape feature in pedestrian safety: A modified multi-level multiple membership model.
Zhu M, Sze NN, Song P … +1 more , Ding H
Accid Anal Prev · 2026 Jul · PMID 42391867 · Publisher ↗
Pedestrians are among the most vulnerable road users in urban transport systems. Studies have explored the effects of the built environment, traffic, and human characteristics on pedestrian crash risk. However, the major...
Pedestrians are among the most vulnerable road users in urban transport systems. Studies have explored the effects of the built environment, traffic, and human characteristics on pedestrian crash risk. However, the majority of these studies have focused on the association between pedestrian safety and influencing factors at macroscopic spatial scales. The influences of micro-level factors, particularly streetscape features, on pedestrian safety are less studied due to the unavailability of high-resolution spatial data. In this study, the influences of streetscape features, in terms of the percentage of green view, sky view, road space, sidewalks, buildings, and traffic signs, on pedestrian crash frequency are evaluated using a computer vision approach and citywide street view imagery. Additionally, the influences of other micro-level factors, such as pedestrian network configuration and transport facilities, and macro-level factors, including land use and socio-demographics, are accounted for. Since data at different spatial scales are used, a modified multi-level multiple membership model is adopted to jointly estimate the influences of both micro- and macro-level factors on pedestrian safety, accounting for possible spatial dependence and between- and within-level correlations. Results indicate that pedestrian crash frequency increases with the presence of a bus stop or metro exit, the number of crosswalks, general and barrier-free walking accessibility, and the percentage of building view. In contrast, pedestrian crash frequency decreases with the presence of a footbridge or underpass and the percentage of green, sky, road space, sidewalk, and traffic sign views at the microscopic spatial scale. Furthermore, land use and population socioeconomics also significantly affect pedestrian safety. The findings should shed light on urban design and planning strategies that can enhance walkability without compromising pedestrian safety.
Assessing autonomous driving performance and environmental influencing factors using real-world operational trajectory data.
Yang H, Yu Q
Accid Anal Prev · 2026 Jul · PMID 42385522 · Publisher ↗
Autonomous vehicles (AVs) are rapidly entering commercial deployment, but large-scale empirical comparisons with human driving remain limited. Using naturalistic data from two demonstration zones in Beijing and Shanghai,...
Autonomous vehicles (AVs) are rapidly entering commercial deployment, but large-scale empirical comparisons with human driving remain limited. Using naturalistic data from two demonstration zones in Beijing and Shanghai, this study analyzes 444 vehicles and over 100 million 1Hz trajectory records from multiple permitted AV operators. A hierarchical segmentation pipeline merges any sub-five-second AV/manual mode-indicator switch with adjacent segments to suppress transient flicker that would otherwise be misread as takeover events, yielding stable segments for multi-granularity comparison of longitudinal dynamics, ride comfort, and human-machine interaction. The two zones represent contrasting operational regimes: a congested Beijing demonstration zone and a higher-speed Shanghai commercial robotaxi corridor, allowing recurrent cross-regime patterns to be separated from regime-sensitive variation. Routine jerk levels are similar between AV and human driving, whereas AV jerk is 47.5% lower during active longitudinal maneuvers in Shanghai. After filtering near-stationary events, in-motion takeovers are classified into Smooth, Braking, and Accelerating types. Accelerating takeovers account for roughly one quarter of in-motion events in both cities, whereas Braking dominates Shanghai and Smooth transitions dominate Beijing. An exposure-adjusted Shapley additive explanations (SHAP) analysis with AV traffic volume as a fixed offset identifies distinct environmental associations: curvature and sinuosity rank highest for Braking, whereas intersection proximity ranks first for Accelerating. Total disengagement counts therefore conflate behaviorally distinct intervention types and city-specific operational variation, both visible only after decomposing events by post-takeover dynamics and stratifying by operating regime.
Multi-scale modeling of electric vehicle fatal crash risk: uncovering spatial heterogeneity and infrastructure-land use coupling mechanisms.
Zhao Y, Wang S, Dong H … +3 more , Xie K, Dong C, Wang M
Accid Anal Prev · 2026 Jul · PMID 42385521 · Publisher ↗
Understanding the spatial heterogeneity of electric vehicle (EV) crash risks is critical for safeguarding electrified mobility systems, yet existing methods cannot simultaneously capture spatial dynamics while maintainin...
Understanding the spatial heterogeneity of electric vehicle (EV) crash risks is critical for safeguarding electrified mobility systems, yet existing methods cannot simultaneously capture spatial dynamics while maintaining model interpretability. By accounting for spatially varying macro-level spatial non-stationarity and complex interaction, a Multi-Scale Adaptive Bandwidth Geographically Weighted Random Forest (MSAB-GWRF) method is developed to uncover the spatial heterogeneity and infrastructure-land use coupling mechanisms of EV fatal crash risk at the regional level. Other than traffic and weather factors, this study investigates the effects of charging infrastructure density, the Social Vulnerability Index (SVI), and land-use characteristics. Meanwhile, a cost-based EV crash risk indicator is formulated by incorporating crash-associated economic costs. Based on monthly state-level observations across the U.S. from 2020 to 2022 (N = 1836), the modelling results indicate that the MSAB-GWRF method outperforms the fixed-bandwidth GWRF, improving average R by 18.2%, particularly in the high-risk regions. More importantly, SHAP-based explanations further reveal pronounced spatial non-stationarity in the drivers of EV fatal crash risk. Land-use and EV-related indicators dominate in most states, outweighing conventional traffic exposure, while social vulnerability dominates in the South and Appalachia. SHAP analysis further identifies a capacity-dependent threshold effect. In high-risk regions, charging infrastructure and land-use intensity act as risk amplifiers once regional exposure exceeds critical levels, whereas low-risk regions absorb comparable expansion without safety deterioration. These findings indicate that uniform safety policies are insufficient, calling for regionally differentiated strategies coordinating charging deployment, land-use planning, and equity-oriented interventions.
Differential sensitivity of self-reported driving and collision measures to aspects of shiftwork, sleep, and fatigue.
Goodwin PJ, Pilkington-Cheney F, Taylor Y … +1 more , Groeger JA
Accid Anal Prev · 2026 Sep · PMID 42378747 · Publisher ↗
Sleepiness, fatigue, and shiftwork are three aspects of daily life that might affect individuals' safety when driving. Driving behaviours are commonly assessed via self-report measures. In this study, different self-repo...
Sleepiness, fatigue, and shiftwork are three aspects of daily life that might affect individuals' safety when driving. Driving behaviours are commonly assessed via self-report measures. In this study, different self-reported measures of driving behaviours are compared to test their sensitivity to measures of sleepiness, fatigue, and night shiftwork. Police shiftworkers from English forces completed a survey that included questions about shiftwork, sleep, fatigue, and included three measures of driving behaviour: The Driver Behaviour Questionnaire (DBQ - Lawton et al., 1997; Reason et al., 1990), Smith's (2016) measures of poor and fatigued driving, and self-reported collision frequencies. Hierarchical regressions confirmed that the driving behaviour measures were differentially sensitive to aspects of sleepiness, fatigue, and shiftwork. Responses to Smith's components of driver fatigue and risk taking, and questions on collision frequencies, but not responses to the DBQ, could be predicted from the proportion of night shifts a person worked. Conversely, only the DBQ was sensitive to fatigue as a predictor of poor driving. All measures were sensitive to aspects of sleepiness and the extent to which it predicted driving outcomes (Adj. R) differed between scales. Self-reported collision data was least sensitive to contexts of shiftwork, sleepiness and fatigue and such data should be used cautiously to explore effects of sleep and shiftwork on driving behaviour. The present findings highlight the importance of using scales that are sensitive to the context of investigations (e.g., shiftwork). Relying on inappropriate scales could lead to the underreporting of poor driving behaviours in sleepy and shiftworking drivers.
Delving into the visual attention of pedestrians during street crossing under time pressure: An eye-tracking approach.
Dhoke A, Choudhary P
Accid Anal Prev · 2026 Sep · PMID 42378746 · Publisher ↗
Pedestrian safety is a major concern, especially in heterogeneous traffic conditions like those commonly seen in India. In such complex environments, how the visual attention is directed to different traffic elements, es...
Pedestrian safety is a major concern, especially in heterogeneous traffic conditions like those commonly seen in India. In such complex environments, how the visual attention is directed to different traffic elements, especially under time pressure, plays a key role in understanding their decision-making and improving their situational awareness. To analyze these visual patterns, the present study examines Average Fixation Duration (AFD) as an indicator of visual attention across several Areas of Interest (AOIs), including two-wheelers, cars, heavy vehicles, signal heads, and the intersection area. Experiments were conducted in a virtual environment that simulated a real-world signalised intersection, using a projector-based pedestrian simulator integrated with an eye-tracker. A total of 62 participants completed crossing trials under three experimentally manipulated time pressure levels: No Time Pressure (NTP), Low Time Pressure (LTP), and High Time Pressure (HTP). The results showed that different levels of time pressure had a clear and significant impact on how pedestrians directed their visual attention toward various Areas of Interest (AOIs). In addition, the study examined other influencing factors, including head-turning behavior before and during crossing, as well as different temporal compliance categories: temporal compliance, non-dangerous temporal non-compliance (TNC), and dangerous TNC. These factors were also found to have a noticeable effect on pedestrians' visual attention patterns. Ultimately, this study provides new insights into pedestrian situational awareness and visual strategies, with practical implications for intersection design.
Differentiating high-frequency and high-severity hotspots: A robust risk-evolution-volume (REV) framework.
Yang S, Liu K, Yue H … +2 more , Jia J, Qiu Y
Accid Anal Prev · 2026 Sep · PMID 42378745 · Publisher ↗
Crash hotspot identification based solely on crash counts or density may over-prioritize high-volume roads and provide limited insight into whether a hotspot is associated with crash frequency or crash severity. This stu...
Crash hotspot identification based solely on crash counts or density may over-prioritize high-volume roads and provide limited insight into whether a hotspot is associated with crash frequency or crash severity. This study proposes a generalizable risk-evolution-volume (REV) framework for differentiating high-frequency and high-severity crash hotspots, with Jiaozhou City, China, used as a case study based on crash data from 2022 to 2024. The framework integrates severity-weighted network kernel density estimation (SW-NKDE), threshold screening based on the cumulative distribution fitted by the Hurdle Gamma model, spatiotemporal evolution analysis, and a two-dimensional relative risk assessment using the critical crash rate (CCR) and a severity-weighted critical crash rate (SWCCR). The results reveal a clear divergence between crash frequency and crash severity patterns under comparable traffic conditions. Urban signalized intersections are more likely to exhibit relatively high crash frequency, whereas suburban segments, suburban unsignalized intersections, and some urban residential-interface locations are more likely to exhibit relatively high crash severity. The results further indicate that locations with seemingly high crash density do not necessarily represent abnormal risk once traffic exposure is taken into account. The S217 provincial highway provides an illustrative example of this pattern. Overall, the proposed framework could be adopted by traffic authorities within annual road safety management programs as a network-level screening tool for identifying hotspots with priority governance value, distinguishing hotspot types, and informing targeted countermeasures and resource allocation.
Modeling takeover decisions in driving automation: a multilevel drift-diffusion model (MDDM) framework integrating human, system, and environmental factors.
Sheng Y, Wu C, Li S … +4 more , Chen Y, Sun X, He D, Zhang J
Accid Anal Prev · 2026 Sep · PMID 42372571 · Publisher ↗
Takeover behaviors represent one of the most critical safety challenges when driving with automation, yet existing models for takeover prediction often overlook the underlying cognitive mechanisms. To address this gap, w...
Takeover behaviors represent one of the most critical safety challenges when driving with automation, yet existing models for takeover prediction often overlook the underlying cognitive mechanisms. To address this gap, we propose a multilevel modeling framework that integrates hierarchical linear modeling with a drift-diffusion model (DDM) to characterize how human, system, and environmental factors jointly shape the cognitive process of takeover decisions. The framework separates rapid fluctuations in environmental influences from stable human and system influences and maps these components onto distinct cognitive parameters. To evaluate the framework, we collected a driving-simulator dataset (N = 64) that systematically manipulated vehicle speed, optical flow, lighting, driving experience, and automation reliability. The fitted model reproduced response-time distributions that highly matched empirical data, indicating high psychological interpretability and modeling validity. These results suggested a dissociable process representation: environmental factors were represented through drift-rate differences, driving experience through decision-boundary differences, and automation reliability through non-decision-time differences. These findings suggest that takeover behavior emerges from distinct cognitive pathways. Moreover, the proposed framework offers a generalizable framework to integrate human, environmental, and system factors into cognitive process models in the context of human-intelligent system interaction.
The state-dependent causal effect of a V2X-based beyond-line-of-sight warning system on young driver response: a causal machine learning approach.
Peng Z, Yu B, Bao S … +3 more , Lu W, Chen Y, Wang J
Accid Anal Prev · 2026 Sep · PMID 42365749 · Publisher ↗
Due to the perceptual challenges created by obstructed sight lines, drivers in beyond-line-of-sight (BLOS) scenarios often fail to detect hazards until insufficient stopping distance remains, making these situations part...
Due to the perceptual challenges created by obstructed sight lines, drivers in beyond-line-of-sight (BLOS) scenarios often fail to detect hazards until insufficient stopping distance remains, making these situations particularly hazardous for young drivers with limited hazard anticipation capabilities. While vehicle-to-everything (V2X)-based warning systems offer a promising countermeasure, precisely quantifying the system effectiveness remains a challenge, as previous studies often rely on associational methods with limited ability to isolate causal effects or characterize treatment effect heterogeneity. This study aimed to apply a causal machine learning framework to estimate the effect of a V2X-based BLOS warning system on young driver response and explore potential sources of treatment effect heterogeneity in a data-driven manner. Based on data from a randomized within-subject driving simulator experiment for causal identification, double/debiased machine learning (DML) was used to estimate the average treatment effect, with the estimates further examined through robustness analyses. The analysis revealed that the BLOS warning had a positive and statistically significant effect on overall response quality. Among trials with identifiable response initiation, the warning was also associated with shorter reaction times. To explore treatment effect heterogeneity, a causal forest implemented within the DML framework was applied to estimate the conditional average treatment effects. An exploratory analysis utilizing Shapley additive explanations (SHAP) suggested that the estimated treatment effects were highly state-dependent. The heterogeneity patterns were more strongly associated with drivers' instantaneous vehicle control states than with other variables, such as static driver characteristics. A subgroup analysis further revealed that the warning benefits were statistically reliable primarily for observations characterized by stable longitudinal control and more active steering correction behavior. These findings highlight the value of incorporating real-time vehicle control states into the design of adaptive advanced driver-assistance systems to improve warning system effectiveness.
How conservative driving behavior increases crash risk: Understanding the systemic safety impacts of older drivers in mixed traffic flows.
Xue K, Jian M
Accid Anal Prev · 2026 Sep · PMID 42364297 · Publisher ↗
Despite the cautious driving behavior of older adults, their over-representation in traffic accidents presents a significant safety concern. Their conservative driving patterns, such as slower reaction times and longer h...
Despite the cautious driving behavior of older adults, their over-representation in traffic accidents presents a significant safety concern. Their conservative driving patterns, such as slower reaction times and longer headways, may disrupt heterogeneous traffic flow and increase crash risks. To investigate this phenomenon, this study proposes a data-driven risk assessment framework that links micro-level car-following behavior with macro-level traffic safety. Using driving simulator data from 16 older and 19 young drivers, a Time Series Lightweight Adaptive Network (TSLANet) is employed to train longitudinal acceleration prediction models for both groups, which significantly outperformed mainstream models such as Intelligent Driver Model (IDM) and Transformer in terms of RMSE, MAE, and R. The framework then enables flexible configurations of older driver proportions and various spatial layouts. It simulates five car-following scenarios and outputs multidimensional indicators, including acceleration noise, speed coefficient of variation (SCV), deceleration rate to avoid a crash (DRAC), and potential index for collision with urgent deceleration (PICUD). Simulation results reveal that increasing the proportion of older drivers amplifies speed fluctuations and reduces safety margins. Front-concentrated and random distributions of older drivers exacerbate risks, whereas rear-concentrated and alternating distributions effectively mitigate adverse effects. The methodology and findings of this study provide robust evidence for refined traffic risk assessment, safety interventions for older drivers, and mixed-traffic optimization in aging societies.
How are pedestrian safety compromised under suppressed warning cyberattacks at a connected intersection? - Exploring vehicle-pedestrian interactions using a hidden markov model-based approach.
Chen C, Li Z, Xu Y … +5 more , Guo J, Wei H, Ahmed MM, Zhang G, Wang K
Accid Anal Prev · 2026 Sep · PMID 42364296 · Publisher ↗
Pedestrians are among the most vulnerable road users at intersections, and their safety risks are expected to decrease while connected vehicle (CV) warning systems can support driver awareness and yielding behavior. Howe...
Pedestrians are among the most vulnerable road users at intersections, and their safety risks are expected to decrease while connected vehicle (CV) warning systems can support driver awareness and yielding behavior. However, cyberattacks that suppress pedestrian warnings may fundamentally alter driver-pedestrian interactions in ways that are difficult to observe in real traffic. This study examines how cyberattacks targeting pedestrian warnings affect pedestrian safety. A controlled driving-simulator experiment was conducted with 32 human drivers to collect real-world vehicle-pedestrian interaction trajectories under both benchmark and cyberattack conditions. In the cyberattack scenario, pedestrian warnings were removed from the dashboard display. To address the limited scale of experimental data and the high variability of pedestrian motion, this study proposes a Hidden Markov Model (HMM)-based generative framework integrating surrogate safety measurements (SSM) to augment the vehicle-pedestrian interaction. Driving simulation analysis shows that cyberattacks can bring significant hazards to vehicle-pedestrian interaction, with a larger reduction in time to collision. At the warning phase, the stop dynamic types' safety is degraded by cyberattacks, with speed increasing over time. The HMM-generated trajectories show that increasing pedestrian speed consistently reduces safety in both benchmark and cyberattack scenarios, with safety measurement being particularly sensitive at low pedestrian speeds (< 0.8 m/s). Under these low-speed conditions, cyberattacks have a more pronounced adverse impact on physically farther pedestrians, especially when the pedestrian-warning distance is greater than 15 m. The results offer insights for the design of pedestrian warning systems and cyber-resilient traffic safety strategies.
HICnet: A two-stage framework based on region partitioning for predicting HIC values from car point clouds.
Li S, Wu J, Shi A … +7 more , Li F, Wang G, Zhou J, Xia Y, Wang L, Li J, Li K
Accid Anal Prev · 2026 Sep · PMID 42349249 · Publisher ↗
Predicting Head Injury Criterion (HIC) values is essential for pedestrian safety assessment but remains challenging due to the complex 3D geometry of vehicles and the long-tailed distribution of HIC data. To capture the...
Predicting Head Injury Criterion (HIC) values is essential for pedestrian safety assessment but remains challenging due to the complex 3D geometry of vehicles and the long-tailed distribution of HIC data. To capture the vehicle's spatial structure, we represent the front-end as a 3D point cloud and introduce HICnet, a framework built upon a PointNet++ backbone for geometric feature encoding, augmented with multilayer perceptrons to integrate auxiliary physical attributes. To address the long-tailed distribution, we devise a dual-pronged strategy grounded in a key observation: HIC values exhibit inherent spatial segregation, with high and low values predominantly clustered in non-hood and hood regions, respectively. Motivated by this spatial heterogeneity, we first explicitly partition the problem, employing dual specialized regression networks for hood and non-hood regions to capture their distinct impact patterns. Second, we design a tailored loss function that incorporates probability density estimation to directly mitigate the data imbalance. Extensive experiments on 14 vehicle models demonstrate that HICnet consistently outperforms baseline architectures, achieving top predictive performance across the majority of test cases while maintaining robustness to diverse vehicle geometries. This offers a reliable and accurate tool for practical automotive safety evaluation.
A macroscopic Game-Theoretic Model Predictive Control framework for emergency vehicle prioritization in urban traffic networks.
Abdi A, Moaveni B, Tettamanti T
Accid Anal Prev · 2026 Sep · PMID 42341637 · Publisher ↗
Efficient Emergency Vehicle (EV) prioritization in congested urban networks remains a critical challenge for smart city infrastructure, as traditional preemption methods often trigger secondary congestion and network ins...
Efficient Emergency Vehicle (EV) prioritization in congested urban networks remains a critical challenge for smart city infrastructure, as traditional preemption methods often trigger secondary congestion and network instability. This paper proposes a decentralized, macroscopic Game-Theoretic Model Predictive Control (GTMPC) framework designed to facilitate rapid EV movement while preserving overall traffic equilibrium. The core of the proposed architecture integrates a weighted non-cooperative game with a potential function structure within a receding-horizon Model Predictive Control (MPC) structure, allowing each intersection to act as an autonomous agent that strategically allocates green-time allocations among competing traffic phases. By assigning dynamic priority weights to EV-serving approaches, the controller proactively clears queues and minimizes stop-and-go behavior. The framework is validated using high-fidelity SUMO simulations initialized with real-world traffic data from Dublin, Ireland, across varying demand profiles, including a 50% flow increase to simulate heavy traffic saturation. Experimental results demonstrate that the GTMPC framework significantly outperforms both Fixed-Time and Classical MPC strategies. Specifically, it achieves substantial reductions in Average Suffering Red-lights, Average Time Loss, and Average Stop Count, while markedly enhancing Average Speed for emergency vehicles. Notably, under heavy traffic conditions, GTMPC maintains robust performance, reducing Average Waiting Time by over 20% in scenarios where standard MPC baselines experience severe congestion and degraded traffic performance. Furthermore, the framework demonstrates an adaptive capability to manage concurrent EV arrivals through strategic multi-player negotiations. Owing to its decentralized structure, the proposed approach shows promising scalability and robustness properties in the conducted simulation studies, making it suitable for large-scale, real-time urban traffic control applications.
Cyclist behavioral persistence across consecutive intersections: Re-ID based analysis and intention prediction.
Li H, Zhang Z, Ren G
Accid Anal Prev · 2026 Sep · PMID 42335682 · Publisher ↗
The safety risks associated with cyclists at signalized intersections are often analyzed within isolated units, an approach that fails to account for the inherent behavioral dependencies across consecutive intersections....
The safety risks associated with cyclists at signalized intersections are often analyzed within isolated units, an approach that fails to account for the inherent behavioral dependencies across consecutive intersections. Ignoring such cross-intersection persistence results in a fragmented understanding of cyclist decision-making and hinders the early identification of risky behaviors. This study investigates cyclists' behavioral persistence across consecutive signalized intersections using high-resolution trajectory data collected from unmanned aerial vehicles and roadside cameras. A re-identification (Re-ID) method is applied to match cyclists across intersections, enabling the analysis of individual-level behavioral persistence. Based on this, a downstream intention prediction framework is developed, incorporating upstream behavioral information as behavioral priors. The results show that cyclists' behaviors are not independent across intersections. While aggregate-level analysis reveals moderate associations, individual-level analysis demonstrates clear persistence in behavior, especially for decision-related variables such as crossing decision and rolling behavior. Incorporating upstream behavioral information significantly improves downstream intention prediction, with the largest improvement observed when individual-level behavioral priors are utilized via Re-ID. The findings highlight the importance of modeling cyclist behavior from a corridor-level perspective. By leveraging cross-intersection behavioral persistence, the proposed framework enables earlier and more reliable identification of risky behaviors, providing practical support for proactive traffic safety management at signalized intersections.
A driving simulator study on drivers' reactions to sudden appearing of a Powered Two-Wheeler in a left-turn manoeuvre.
Carini C, Paliotto A, Meocci M … +3 more , Mazzi C, Torre F, Marradi A
Accid Anal Prev · 2026 Sep · PMID 42335681 · Publisher ↗
Nowadays, it is widely recognised that most accidents are caused by human errors and many of them are related to the delay in reacting to sudden stimuli. Even small delays in reaction time can have substantial implicatio...
Nowadays, it is widely recognised that most accidents are caused by human errors and many of them are related to the delay in reacting to sudden stimuli. Even small delays in reaction time can have substantial implications, as they directly affect the stopping distance and the ability to avoid potential hazards. The driving simulator study carried out aims at investigating the reaction time and type of drivers in a critical scenario. The scenario considered is a T-intersection where the ego vehicle is driving on the main road and must turn left at the intersection. While the vehicle is turning, a Powered Two-Wheeler (PTW) travelling in the opposite direction suddenly appears and crosses the trajectory of the ego vehicle. Eye-tracking glasses were used to determine where the driver's gaze was directed throughout the simulation. Many different parameters have been studied in the experiment. The relationship between the 70 participants' characteristics (gender, age, experience) has been related to the reaction times and reaction types (breaking or steering). Additional parameters have been also studied considering their influence on reaction time, such as the position of the gaze in the field of view when the PTW appears. The results show an average response time of about 0.9 s from the moment the PTW appears and the start of the reaction, with a large predominance of steering manoeuvres. Such results can be also useful in the development of in-vehicle systems able to inform the driver about a danger or autonomously react to it.
Game-theoretic modeling of vehicle unprotected left turns considering drivers' bounded rationality.
Lian Y, Zhang K, Li S … +1 more , Li M
Accid Anal Prev · 2026 Sep · PMID 42335680 · Publisher ↗
Unprotected left turns at signalized intersections represent one of the most critical safety challenges in urban traffic, contributing significantly to severe accidents due to the complex interactions between conflicting...
Unprotected left turns at signalized intersections represent one of the most critical safety challenges in urban traffic, contributing significantly to severe accidents due to the complex interactions between conflicting vehicles. Given the interactive nature of these conflicts, game theory serves as an ideal framework for modeling vehicle decision-making in unprotected left turns. Although some studies have advanced this by introducing bounded rationality to account for human cognitive limitations, they largely depend on static or heuristic rationality parameters. Such approaches are insufficient to capture the dynamic, interaction-aware evolution of human cognition during complex unprotected left-turn maneuvers. To address this gap, we propose a novel decision-making model for vehicle unprotected left-turn scenarios that integrates game theory with explicit considerations for drivers' dynamic bounded rationality and decision tendency. Our model is formulated as a two-player normal-form game solved by a quantal response equilibrium (QRE), offering a probabilistic depiction of driver decision-making processes that accounts for driving styles and dynamic risk-taking behaviors. We introduce a neural-embedded expectation-maximization (EM)-like alternating optimization framework to calibrate interaction-aware bounded rationality parameters, utilizing high-fidelity microscopic trajectory data for empirical grounding. Simulation experiments demonstrate that the proposed model captures human decision tendencies in high risk scenarios more accurately than perfectly rational models. By quantifying the degree of rationality in critical conflict zones, these findings provide essential insights for developing autonomous driving systems capable of anticipating human error and preventing collisions in mixed traffic environments.
Uncovering pedestrian stress and crossing strategies via mixed-reality encounters with autonomous vehicles.
Liu Y, Ji Y, Feng T
Accid Anal Prev · 2026 Sep · PMID 42330781 · Publisher ↗
Pedestrian crossings when encountering autonomous vehicles (AVs) is a complex interplay between vehicle kinematics, external human-machine interface (eHMI) cues, and pedestrians' physiological states. However, studies ha...
Pedestrian crossings when encountering autonomous vehicles (AVs) is a complex interplay between vehicle kinematics, external human-machine interface (eHMI) cues, and pedestrians' physiological states. However, studies have not yet integrated physiological arousal with observed crossing behaviour in AV settings, nor clarified how stress-related responses differ across pedestrian groups. In this study, we investigate how AV traffic scenarios (e.g., vehicle speed, temporal gaps, yielding behaviours, urgency prompts, and eHMI) shape physiological arousal which affects crossing decisions, aiming to explain both direct effects and arousal-mediated mechanisms to support safer eHMI and road design. A head-mounted mixed reality (MR) experiment combined with galvanic skin response (GSR) measurement was developed to project life-size virtual AVs into real streetscapes and to quantify sympathetic arousal. Participants were stratified into three emergency reactivity groups (low/medium/high) using the Emergency Reaction Questionnaire (ERQ) traits. A multi-group path model was adopted to identify how crossing behaviour is influenced directly and indirectly by traffic conditions through physiological arousal across different groups. Results indicate that across groups, roadway cues and eHMIs signals generally have direct effects and arousal-mediated effects on crossing behaviour, but such mediation is mainly observed in medium and high emergency reactivity groups. Under high-intensity cue conditions, salient stimuli increase sympathetic activation, which can prolong pre-step-off waiting and, under strong cues, may also attenuate or offset the net change in walking speed via arousal-mediated pathways, depending on emergency reactivity groups. The strength of mediation and behavioural elasticity varies considerably across different groups. These insights help formulate group targeted safety interventions, thus providing actionable guidance for the system design involving AV and pedestrian interactions.
Toward safe and efficient autonomous platooning: A coalitional game-theoretic framework.
Wei Z, Wang H, Yang K … +3 more , Xu Z, Yip WS, To S
Accid Anal Prev · 2026 Sep · PMID 42322942 · Publisher ↗
Autonomous vehicle platooning in mixed-traffic environments faces critical safety challenges due to uncertain interactions with surrounding vehicles and the limitations of rigid platoon structures. In practice, unpredict...
Autonomous vehicle platooning in mixed-traffic environments faces critical safety challenges due to uncertain interactions with surrounding vehicles and the limitations of rigid platoon structures. In practice, unpredictable behaviors of human-driven vehicles can disrupt coordination and induce disturbances, while fixed platoon organizations lack the flexibility to adapt to evolving traffic conditions, leading to increased collision risk and degraded safety margins. To address this issue, this paper proposes a safety-oriented hierarchical framework for autonomous platooning, featuring a centralized decision-making and distributed execution architecture. At the upper level, a coalitional game-theoretic decision-making mechanism is developed to enable connected autonomous vehicles to form adaptive coalitions and generate interaction-aware behavioral strategies that explicitly consider both internal platoon safety and external traffic interactions. At the lower level, distributed motion controllers translate these strategies into dynamically feasible control actions, ensuring safe execution under vehicle constraints. The proposed framework is validated in representative mixed-traffic scenarios. Quantitative results demonstrate that the method significantly improves safety performance by increasing minimum time-to-collision and reducing exposure to critical risk conditions, while maintaining comparable operational efficiency. These findings confirm the effectiveness of the proposed framework in enabling safe and efficient autonomous platooning in complex traffic environments.
Safety evaluation of chevron markers and speed reduction markings for mountainous freeway combined alignments: a driving simulator study.
Fu H, Wang X, Hurwitz DS … +4 more , Ye X, Lei J, Ye J, Zhou R
Accid Anal Prev · 2026 Sep · PMID 42322941 · Publisher ↗
Mountainous freeways generally deploy safety countermeasures on crash-prone sections to mitigate crash risks from complex combined horizontal-vertical alignments, highlighting the importance of countermeasure evaluation...
Mountainous freeways generally deploy safety countermeasures on crash-prone sections to mitigate crash risks from complex combined horizontal-vertical alignments, highlighting the importance of countermeasure evaluation in enhancing traffic safety. While international practices regarding safety improvements for crash-prone roadways have developed quantitative evaluation for implemented countermeasures, the specific design schemes of proposed countermeasures in the design stage usually rely on engineers combining engineering experience to make reasonable inferences based on design specifications, supplemented by crash features and cost-effectiveness. However, it is essential to validate the rationality of such proposed countermeasure design configurations (specifically for countermeasures involving complex driver-geometry interactions) before implementation, to provide quantitative references for engineers to refine their schemes. Driving simulation is a powerful tool for this purpose, as it complements engineering inferences with behavioral-level quantification to validate and refine design schemes. Hence, this study developed a driving simulation-based framework for pre-implementation evaluation of chevron alignment markers and longitudinal speed reduction markings on seven combined alignment types within a 35 km crash-prone section of an operational mountainous freeway in China. Using the high-fidelity Tongji University Driving Simulator, experiments involving 30 participants were conducted to collect vehicle operational data. Paired Wilcoxon signed-rank tests quantitatively evaluated effectiveness using five surrogate safety measures. Results revealed significant section-specific effectiveness: (1) Both countermeasures were most effective on curve-downgrade sections; (2) Neither significantly improved safety on curve-crest or curve-upgrade sections; (3) Both countermeasures were particularly effective for sections with R < 2500 m, G < |2.0| %, and ΔG ≥ |2.0| %. These findings provide quantitative guidance for engineers to refine section-specific design schemes and resource allocation during the planning stage, ensuring safety countermeasures customized for the unique demands of different combined sections.
A Misclassification Framework for Identifying Systematic Error in Pedestrian Crash Prediction: Evidence from Seattle, WA.
Douglas G, Mooney SJ, Prendez DM … +1 more , Boyle LN
Accid Anal Prev · 2026 Sep · PMID 42320389 · Publisher ↗
Pedestrian crash prediction models often fail systematically, yet distinguishing correctable failures from random errors is challenging. We present a framework combining spatial clustering, temporal persistence, and mode...
Pedestrian crash prediction models often fail systematically, yet distinguishing correctable failures from random errors is challenging. We present a framework combining spatial clustering, temporal persistence, and model ensemble consensus to identify improvement opportunities. Applied to 13,706 Seattle intersections, the framework identified 141 high-priority locations (1.03%) exhibiting elevated temporal persistence, strong spatial clustering, and high ensemble consensus. A multiplicative priority score serves as a proxy for systematic error, concentrating 6.6% of total error into 1.03% of locations. The framework flagged crash-enriched intersections (roughly twice the baseline crash rate) where all models over-predicted risk, indicating the feature set captures that a location is hazardous but lacks the site-level descriptors needed to calibrate how much. Audits revealed three categories of omitted features: visual obstructions, static road use, and network complexity. Validation confirmed that adding these features reduced priority scores by 6.1% at high-priority locations, confirming that the framework correctly identifies where omitted context matters most.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe