Analytic Methods in Accident Research最新文献

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

Recently released autonomous vehicle datasets like Waymo can provide rich information (and unprecedented opportunities) to investigate lane-changing behaviour of autonomous vehicles, requiring data from multiple drivers and lanes with different objectives. As such, the study investigates the discretionary lane-changing execution behaviour of autonomous vehicles and compares its behaviour with human-driven vehicles from Waymo and Next Generation Simulation (NGSIM) datasets. Several behavioural factors are statistically analysed and compared, whereas the discretionary lane-changing execution time (or duration) is modelled by a random parameters hazard-based duration modelling approach, which accounts for unobserved heterogeneity. Descriptive analyses suggest that autonomous vehicles maintain larger lead and lag gaps, longer discretionary lane-changing execution time, and lower acceleration variation than human-driven vehicles. The random parameters duration model reveals heterogeneity in discretionary lane-changing execution behaviour, which is higher in human-driven vehicles but decreases significantly for autonomous vehicles. Whilst contradictory to a general hypothesis in the literature that autonomous vehicles will eliminate heterogeneity, our finding indicates that heterogeneous behaviour also exists in autonomous vehicles (although to a lesser extent than in human-driven vehicles), which can be contextual to prevailing traffic conditions. Overall, autonomous vehicles show safer discretionary lane-changing behaviour compared to human-driven vehicles.

Traffic conflicts are generally considered independent events in existing extreme value theory models to estimate the risk of total or single types of crashes. However, traffic events at a road entity are not necessarily independent interactions and can lead to multiple traffic conflicts with shared common unobserved factors. A comprehensive estimation of crash risks in a road entity needs to consider the correlation of potential traffic conflicts to avoid possible bias in prediction performance and the problem of undetected deficiencies. This study proposes a Bayesian non-stationary bivariate generalised extreme value modelling framework to estimate the severe and non-severe crash risks accounting for the correlation between right-turn and rear-end conflicts at signalised intersections. A deep neural network-based computer vision technique was applied to extract the traffic conflicts from 77 h of video recordings over two right-turn approaches at two signalised intersections in Cairns, Australia. Post encroachment time and modified time to collision were used to characterise right-turn and rear-end conflicts, respectively, while an expected post-collision velocity difference was combined with post encroachment time and modified time to collision for crash risk estimation by injury severity levels. Several covariates were used to address the time-varying heterogeneity of traffic conflict extremes and to estimate the differential crash risks at signal cycles. Results showed a significant correlation between right-turn and rear-end crashes at signal cycle levels, indicating the importance of accounting for the dependency among traffic conflict types. Overall, the bivariate models considering the correlation among traffic conflict types were found to understandably perform better than their univariate counterparts. This study provides a demonstration of a correlated crash risk modelling framework that addresses issues related to the suitable traffic conflict measures, time varying risks (non-stationarity), heterogeneity, and injury severity levels of different crash types.

This paper presents an error components mixed logit with heterogeneity in means and variance to capture the heterogeneous effects of contributing factors on fixed object occupant severity. One year (2021) of crash data on fixed object related crashes in Lubbock County, Texas was analyzed with fixed object details extracted from crash narratives and classified into 11 groupings. Crash data included any fixed object collision occurring at any point in the sequence of crash events (not exclusive to the first harmful event). The random parameters were identified as indicators for occupant involvement in the first harmful crash sequence event, with that event being collision with a fixed object, for possible injury and injury severity outcomes. Heterogeneity in the means of these random parameters was found with respect to six different indicator variables. Additionally, heterogeneity in the variance of the injury random parameter was found with respect to two different indicator variables. Inclusion of two error component nests improved prediction accuracy at the observation level for higher severity outcomes. The findings in this study suggest that fixed object classification types should be explored further in relation to heterogeneous effects on occupant severity outcomes. Furthermore, the findings also highlight the applicability of an error components mixed logit model for severity analysis.

The current research effort contributes to safety literature by developing an integrated framework that allows for the influence of independent variables from crash type and severity components at the disaggregate level to be incorporated within the aggregate level propensity to estimate crash frequency by crash type and severity. The empirical analysis is based on the crash data drawn from the city of Orlando, Florida for the year 2019. The disaggregate level analysis uses 15,518 crash records of three crash types including rear end, angular and sideswipe. Each crash record contains crash specific factors, driver and vehicle factors, roadway attributes, road environmental and weather information. For aggregate level model analysis, the study aggregates the crash records by crash type over 300 traffic analysis zones. An exhaustive set of independent variables including roadway and traffic characteristics, land-use attributes, built environment and sociodemographic factors are considered in this level. The empirical analysis is further augmented by employing several goodness of fit and predictive measures. A validation exercise is also conducted using a holdout sample to highlight the superiority of the proposed integrated model relative to the non-integrated model system. The proposed framework can also incorporate unobserved heterogeneity in the model system. The findings of the study indicate that the proposed framework is advantageous for capturing the variable effects simultaneously across the aggregate and disaggregate levels.

Non-motorist injury severity can be affected by various observed and unobserved attributes related to the crash location type (segment or intersection). Recognizing the distinct non-motorist injury severity profiles by crash location type, we propose a joint modeling framework to study crash location type and non-motorist injury severity as two dimensions of the severity process. We employ a copula-based joint framework that ties the crash location type (represented as a binary logit model) and injury severity (represented as a generalized ordered logit model) through a closed form flexible dependency structure to study the injury severity process. The proposed approach also accommodates the potential heterogeneity (across non-motorists) in the dependency structure. The data for our analysis is drawn from the Central Florida region for the years of 2015 to 2021. The model system explicitly accounts for temporal heterogeneity across the two dimensions. A comprehensive set of independent variables including non-motorist user characteristics, driver and vehicle characteristics, roadway attributes, weather and environmental factors, temporal and socio-demographic factors are considered for the analysis. We also conducted an elasticity analysis to show the actual magnitude of the independent variables on non-motorist injury severity for the two locations. The results highlight the importance of examining the effect of various independent variables on non-motorist injury severity outcome by crash location type.

Rear-end crashes particularly on freeways are the most frequent type of collisions causing many injuries, damage and congestion. This paper investigates the impact of varying speed differences between following and leading vehicles on injury severity in two-vehicle rear-end crashes. It develops three groups of correlated joint random parameters bivariate probit models with heterogeneity in means. The rear-end crash data from 2019 to 2021 on Interstate freeways in Florida are utilized, and categorized into periods before, during, and after the COVID-19 pandemic. The study considers two potential injury severity outcomes: no injury and injury/fatality, for both drivers involved in these crashes. The findings indicate that a range of variables, including driver, vehicle, roadway, environmental, crash, and temporal attributes, significantly influence the injury severity outcomes for drivers in both following and leading vehicles. Demonstrating superior goodness-of-fit, the proposed approach sheds light on interactive unobserved heterogeneity, captured through heterogeneity in means and significant correlations among random parameters. The study observes critical influences on the injury severity outcomes of both drivers, with significant factors such as gender, age, vehicle type, weather conditions, lighting, and time of day. Furthermore, the results substantiate the heightened risk outcomes associated with greater speed differences and the period of the COVID-19 pandemic. These findings yield further insights into the risk mechanisms of two-vehicle rear-end crashes and offer guidance for the development of effective safety countermeasures.

In this study, we propose a novel integrated parametric framework for analyzing multivariate crash count data based on linking a univariate count model for the total count of motor vehicle crashes across all possible crash states with a discrete choice model for crash event state given a crash. In doing so, we are able to use information at the disaggregate crash-level from an unordered model structure in analyzing the aggregate level crash count. To our knowledge, this is the first such model proposed in the econometric literature. We apply this approach in a demonstration exercise to examine the number of motor vehicle crashes in Census Block Groups (CBGs) in Austin, Texas, considering four injury severity levels. At the disaggregate level, we incorporate several explanatory variables such as the characteristics of the most severely injured individual and at-fault vehicle’s parties, crash time variables (time of day, weather), crash location variables, and CBG level variables. At the aggregate level, we consider CBG level variables, including road design factors, land-use variables, crash exposure factors, aggregate sociodemographic attributes, and crime and traffic violations related measures. Importantly, our results indicate a significant and positive linkage between the disaggregate crash event state dimensions and the total crash count. Through the use of elasticity measures, our results also clearly highlight the improved policy sensitivity of the integrated model framework.

Traffic safety around schools is a major concern for policy makers and as such safety interventions are often targeted near schools. This paper shows the importance of accounting for the potential endogeneity of proximity to school when attempting to estimate its impact on traffic safety. In this research, we use a Bayesian simultaneous econometric approach with heterogeneity in covariance to disentangle the true effect of proximity to school on cyclist injury frequencies at signalised intersections in an urban setting. We assess the robustness of the bivariate normal assumption, using a scale mixing approach. Notably, we found that proximity to school was associated with an increase in cyclist injuries and this association was stronger when endogeneity was accounted for in the model, confirming the importance of considering endogeneity in studies of traffic safety near schools. Our heterogeneity in covariance specification revealed systematic variations in the covariance structure, which would otherwise go unobserved, providing further insights into sources of heterogeneity with the same set of variables available in the data. A safety-in-numbers effect is also found for cyclists in the study area and period. This research offers policy implications based on the findings of the analysis including the need for safety interventions at intersections with high vehicle turning counts and those in proximity to public transport stops, and better informing decision-makers regarding the magnitude of the impact of proximity to school on cyclist safety at intersections.