Despite the significant efforts made by the researchers and policy-makers, each year almost 1.35 million people lose their lives due to traffic accidents, and the time lag between the accident happening and the emergency services’ awareness of it stands out as the weakest point of the entire pipeline. Typically, a conventional detection pipeline consists of a convolutional feature extractor followed by a temporal classifier. Although such systems work well, they are too computationally intensive to be deployed to roadside cameras due to their hardware limitations. This paper proposes another solution: temporally sorted grayscale video frames are supplied to the Long Short-Term Memory network without using any convolutional network as a front-end. The model contains approximately 9.5 million parameters, is trained on a relatively small amount of annotated data, and has an inference speed that is sufficient for continuous surveillance of CCTV cameras. The experiments have been done on the complete Car Crash Dataset, which includes 1500 videos of traffic accidents and 3000 videos of normal driving conditions from dashboard cameras. A stratified five-fold cross-validation protocol has been adopted for robustness evaluation. In terms of test sets, the system has shown an overall accuracy of 87.5%, 100% recall for accident events, 82.35% precision, and 90.32% F1-score. The last metric is crucial for ensuring the proper functioning of the system in critical situations, where a single failure can result in a life-threatening consequence. The system has been benchmarked on commodity CPU hardware which confirmed the real-time capability of per-clip inference.
Environmental challenges associated with climate change and pollution have intensified worldwide, and the transportation sector is widely recognized as one of the key sources of these problems. Consequently, electric vehicles (EVs) are increasingly viewed as an important alternative for mitigating environmental impacts and promoting sustainable mobility. This study integrates the Stimulus-Organism-Response framework with insights from the Technology Acceptance Model and Unified Theory of Acceptance and Use of Technology to examine factors associated with attitude and trust, and their subsequent relationships with purchase intention toward EVs among prospective urban adopters in Vietnam. The proposed model classifies influencing factors into three categories: benefit-related (perceived usefulness, perceived ease of use, perceived value), personal (environmental concern, personal norms, personal innovativeness), and situational (social influence, vehicle performance, facilitating conditions). Data from 438 respondents were collected using a structured questionnaire and analyzed with Partial Least Squares Structural Equation Modeling. The results reveal that all factors significantly influence attitude and trust, which significantly predict purchase intention. Mediation analysis further indicates that attitude significantly mediates the effects of perceived usefulness, environmental concern, social influence, and vehicle performance on purchase intention, whereas no significant indirect effects are transmitted through trust. Notably, social influence was the strongest predictor of consumer attitude, while personal innovativeness showed a strong positive effect on consumer trust. These findings highlight differentiated psychological pathways through which functional, personal, and situational factors contribute to EV purchase intention and provide practical implications for strengthening social acceptance, communicating consumer value, reducing uncertainty, and developing supportive infrastructure.
Time-to-collision is a commonly employed measure for rear-end collision prediction. However, its conventional formulation, which assumes constant speed and heading, is incapable of identifying sideswipe collisions. A two-dimensional extension has been proposed to incorporate lateral interactions with passenger cars, yet it assumes identical, fixed headings and does not accommodate articulated vehicles such as tractor-semitrailers. In this paper, the existing formulation for the car is first refined to incorporate differences in vehicle heading. Subsequently, new two-dimensional time-to-collision measures are proposed for articulated vehicles: TTC2DAV and modified TTC2DAV. These measures employ constant-speed and constant-acceleration assumptions and are analogous to their one-dimensional counterparts: time-to-collision and modified time-to-collision. The proposed measures are assessed in CARLA using randomly generated cut-in scenarios simulated with a tractor-semitrailer model, incorporating a range of trailer lengths. A short analysis is also conducted to test the measures in a roundabout and tight turn. The analyses demonstrate that the proposed measures substantially improve the detection of sideswipe collisions while maintaining a comparable level of performance to existing measures in detecting rear-end collisions. Across 30 simulated scenarios, they correctly identify 14 of 15 sideswipe collisions, compared with 7 of 15 identified by the existing formulation. Moreover, the mean prediction error for sideswipe collisions is reduced by approximately 20% compared to the existing formulation.
Background Paramedics work in high-risk environments during emergency response and patient transportation. Despite seatbelt policies and other safety regulations, safety varies due to operational demands, workplace culture, and systemic barriers. Previous studies have primarily quantified paramedic injury rates, but few have explored the underlying human factors and decision-making that influence safety behaviours. Aim This study aimed to explore paramedics’ experiences and perceptions of safety during emergency response and transportation, using the framework method guided by the Human Factors Analysis and Classification System (HFACS) systemic framework. Methodology Focus groups were conducted with paramedics in Australia. Deductive coding was conducted collaboratively to ensure rigour and reliability. Analysis of the codes was completed using the HFACS framework. Results Analysis of the transcripts from the focus groups with Australian paramedics identified systemic and cultural factors that normalised unsafe practices. Organisational culture strongly influenced safety behaviour, with junior staff emulating senior colleagues and possibly feeling unable to challenge unsafe norms. Policies were often described as outdated, unclear, or punitive, eroding legitimacy. Inadequate driver and safety training, coupled with ergonomic barriers such as poorly designed restraint systems, further constrained safe practice. These interrelated influences created conditions in which unrestrained travel and other risk-taking behaviours became routine, reflecting a broader normalisation of risk within paramedic culture. Conclusion The study highlights systemic, environmental, and behavioural factors influencing paramedic safety. Participants described unclear or outdated policies, hierarchical pressures, inadequate driver and safety training, and poorly designed restraint systems, all of which constrained safe practice and reinforced unrestrained transportation practices.
Trust is a critical determinant of public acceptance of automated vehicles, particularly in shared mobility contexts where passengers co-experience automation. While prior research has largely examined trust as an individual judgement, emerging shared automated mobility raises the need to understand how trust unfolds within groups and across different sources of driving agency. This study investigates how situational urgency and driving agency (human vs. automated control) shape trust during shared rides. Thirty-six participants completed a within-subjects virtual reality (VR) study in triads, experiencing automated and manual driving under low- and high-urgency scenarios. Trust was assessed using a multi-method approach combining real-time behavioural signals (press-and-hold trust-loss input), post-trial questionnaires, and thematic analysis of group conversations. Results show that trust was lowest in automated low-urgency scenarios and highest in manual high-urgency scenarios, suggesting that perceived agency strongly shapes trust calibration. Button-press behaviour captured moment-to-moment trust breakdowns and correlated with subjective trust ratings, supporting its validity as a real-time indicator. Qualitative findings reveal that trust was socially situated, with group sense-making processes shaping how passengers interpreted vehicle behaviour, situational ambiguity, and control attribution. Together, the findings position trust in automated mobility as a socially embedded and context-dependent phenomenon, and highlight the importance of designing shared automated vehicle (AV) systems that support real-time trust expression and collective understanding.
Having anxiety, depression, or bipolar disorder can significantly impact one’s ability to drive. However, the specific effects of these conditions on drivers’ psychophysical capacities and the likelihood of accidents are not always well documented in the scientific literature. The aim of this study is to examine the relationship between mood disorders, specifically depression, anxiety, and bipolar disorder, and driving. A systematic review on this topic was conducted and registered with PROSPERO (CRD420251023655). A systematic search was carried out in 2025 across seven scientific databases following PRISMA guidelines. Two groups of keywords were used, one for psychopathological disorders (in the title) and another for driving-related terms (in title, abstract, and keywords), adapting them to each database’s search syntax. Filters limited results to peer-reviewed reviews published in English between 2015 and 2025. On-road studies indicate that major depression, more than mild symptoms, is associated with impairments in executive functions critical for driving, while ongoing pharmacological treatment can mitigate some negative effects. Simulator studies suggest that drivers with partially remitted depression can perform comparably to healthy drivers, underscoring the importance of maintaining psychosocial functionality. Anxiety undermines self-efficacy and attentional control, particularly affecting performance in older drivers. Bipolar disorder shows persistent deficits in sustained attention and information processing, even during remission phases. In professional drivers, multiple studies show how work-related stress, fatigue, road rage, and lack of organizational support directly influence risky behaviors. Personality traits like neuroticism and impulsivity, along with mental health disorders, significantly increase the risk of driving errors and traffic violations.
This study aimed to estimate cost parity between electric two-wheelers (E2W) and internal combustion engine (ICE) motorcycles in Indonesia using a segment-specific Total Cost of Ownership model expressed as Equivalent Uniform Annual Cost (EUAC). A study-specific comparative-parity procedure was used to match representative vehicles in the Low, Low-Mid, and Mid-High segments. The deterministic engine evaluated 5,670 technical cases across mileage, ownership horizon, E2W residual value, battery cycle life, and gasoline price. Applying six purchase-subsidy levels expanded the cases to 34,020 paired ICE–E2W comparisons. The results showed that across the full grid, E2Ws had lower EUAC in 81.7 % of Low, 77.6 % of Low-Mid, and 78.7 % of Mid-High comparisons. Without purchase subsidy, the corresponding shares were 52.2 %, 50.1 %, and 54.0 %. Meanwhile, at a Rp7 million subsidy, the shares increased to 89.5 %, 85.3 %, and 85.8 %. In the five-year, 40 km/day reference case with 100 % baseline resale and 800-cycle life, the subsidy required for parity at gasoline Rp10,000/liter was Rp2.37 million, Rp3.11 million, and Rp2.49 million for the Low, Low-Mid, and Mid-High segments, respectively. Under the same assumptions, the required subsidy fell to zero when gasoline price reached Rp15,000–20,000/liter. The E2W resale equation was based on 25 listings and yielded a leave-one-out mean absolute percentage error (MAPE) of 13.8 %. Therefore, long-horizon results were interpreted as conditional scenario estimates rather than precise forecasts. Battery swapping reduced owner exposure to replacement shocks, while fixed-battery outcomes remained sensitive to replacement timing and residual value. Diffusion and loss-aversion theories were used only to interpret the cost results, as consumer choice and adoption were not modeled directly.
Cities are increasing investments in walkable environments due to the health, environmental, and social benefits of walking. However, accurately measuring pedestrian patterns remains challenging due to data limitations. Crowdsourced data presents an opportunity to address data limitations but there are also challenges. This paper aims to take a systematic look into crowdsourced data that are being used to fill gaps and advance pedestrian research with three primary objectives: (1) to identify key gaps and opportunities in the current application of crowdsourced data in pedestrian-related research, (2) to critically evaluate the methods used to process and integrate crowdsourced data for pedestrian research, and (3) to provide key recommendations to guide future research and planning efforts through the effective use of crowdsourced data in pedestrian studies. Conducting a systematic literature review, we screened an initial set of 653 papers to identify 48 peer-reviewed studies published between 2011 and 2024 for full review. We identify eight thematic uses of crowdsourced data in pedestrian research and planning: travel patterns, infrastructure detection, mode detection, navigation, policy, route/infrastructure choice, safety, and walkability/accessibility. Additionally, we highlight four promising future research opportunities: understanding disruptions that shape pedestrian mobility, incorporating equity considerations into research and planning, integrating pedestrian perspectives into crowdsourced imagery, and refining methodologies for analyzing crowdsourced data. We provide a guide for both researchers and policy makers in advancing pedestrian research through the integration of crowdsourced and traditional data sources.
In the current context, in which climate change is a major challenge, the tourism sector can play an important role in mitigating its effects with actions that involve transitioning towards more sustainable mobility. Renting an electric vehicle is an option that is particularly valuable in island territories. Especially on those islands that have fragile ecosystems and high dependence on fossil fuels, which is the case of the Canary Islands. This research aims to identify, using discrete choice models, the factors that influence tourists’ willingness to pay to rent an electric vehicle when visiting the island of Tenerife. This is a relevant issue given the scarcity of data and studies that analyse this issue, especially in island territories. Consequently, a questionnaire was designed to gather data on tourists’ socio-demographic characteristics, travel preferences and other issues relating to vehicle rental. Among the main results, it is worth noting that the tourists most likely to be willing to pay to rent an electric vehicle are young people under 30 years of age with medium-high incomes, who consider that vehicles’ consumption and emissions are important, among other characteristics. Knowledge of these tourists’ profile provides valuable information for policymakers and car rental companies that could be used to better segment the market and develop specific strategies to promote more sustainable mobility on Tenerife.
The introduction of public transport infrastructures is typically accompanied by political discourse expecting “shock effects” on the supply side to translate unconditionally into sustainable mobility practices on the demand side. This study examines how mobility practices have evolved in the wake of such a shock effect, focusing on the introduction of the Léman Express, a cross-border regional rail network that substantially enhanced public transport in the Greater Geneva. Drawing on two surveys conducted before (2018) and after (2022) the inauguration, we combine inductive Multiple Correspondence Analysis with Difference-in-Differences design and Propensity Score Matching to compare matched populations exposed and not exposed to the new infrastructure. We identify two latent dimensions structuring mobility practices in the region – environmental concern and valuation logic – along which the evolution of practices can be tracked while controlling for the confounding role of the COVID-19 pandemic. Our findings reveal that exposure to the Léman Express is associated with a partial attenuation of a general decline in sustainable mobility practices over the period, rather than with a positive shift. This attenuation is moreover narrow in scope – concentrated on environmental concern – and unevenly distributed across the population, remaining confined to urban, well-educated subgroups. By holding the behavioral, attitudinal, and practical aspects of mobility within a single analytical frame, this study contributes to an interdisciplinary dialogue between Transport Studies, Transit Oriented Development scholarship, and Social Practice Theories, and offers evidence on how mobility practices (do not) adapt when confronted with a major infrastructural change.
Jaywalking is widely studied as a compliance problem, yet what actually happens during the jaywalking process remains understudied. This study deploys roadside LiDAR at two urban intersections in Munich, Germany to decompose jaywalking into discrete kinematic states, examining temporal patterns of red-light violation, spatial adaptation under restricted sightlines, and behavioral responses to hazardous encounters. Red-light violations clustered at two ends of the red phase with systematically different walking speeds, suggesting structured temporal adaptation to signal-cycle dynamics rather than random noncompliance. Notably, at the unsignalized intersection where parked vehicles obstruct sightlines, pedestrians stood nearly 2.4 times farther from the curb (median 1.86 m versus 0.78 m, p < 0.001), and the proportion of crossings involving stopping and spatial repositioning was substantially higher than at the signalized site. Both patterns are consistent with information-seeking behavior under conditions of limited traffic information. Concurrently, the hourly rate of hazardous encounters was approximately four times higher at the unsignalized site, and the odds of stopping were 4.7 times as high in car encounters as in cyclist encounters, indicating that behavioral responses during jaywalking vary with threat type. The findings characterize jaywalking as a structured, context-dependent process and demonstrate how high-resolution naturalistic trajectories can support intersection design, pedestrian safety assessment, and autonomous vehicle development.
This study develops a real-time multi-objective path optimization framework for green logistics using the Honey Bee Algorithm (HBA) to address the limitations of conventional routing approaches under dynamic traffic and environmental conditions. The proposed framework simultaneously optimizes fuel consumption, carbon emissions, and delivery time, thereby integrating operational efficiency with environmental sustainability. A simulation-based experimental design is employed to compare HBA with three established metaheuristic algorithms—Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Ant Colony Optimization (ACO)—under controlled logistics scenarios. The results demonstrate that HBA consistently outperforms the benchmark algorithms, reducing fuel consumption by approximately 20% and carbon emissions by approximately 25%, while achieving on-time delivery performance of up to 95%. These results indicate that the adaptive exploration–exploitation mechanism of HBA enables more effective route selection under dynamic and uncertain transportation conditions. The findings further demonstrate that incorporating environmental objectives directly into routing decisions can simultaneously improve operational efficiency and sustainability performance without imposing impractical computational demands. The proposed framework therefore provides a scalable decision-support approach for logistics firms, particularly in urban and last-mile delivery environments where traffic conditions and delivery requirements change dynamically. Overall, the study contributes empirical evidence on the effectiveness of bio-inspired optimization for sustainability-oriented logistics and demonstrates the potential of real-time multi-objective routing to support simultaneous improvements in transportation efficiency and environmental performance.
Electric vehicles (EVs) are increasingly being adopted around the world, positioning them as key components in the decarbonization of the automotive sector. However, forecasting EV demand remains a significant global challenge due to the complexity of consumer behavior and market dynamics. This study presents a hybrid approach that combines agent-based modeling (ABM) with nonlinear diffusion models, specifically the Bass diffusion model, to project the adoption of EVs. The agent-based model simulates the individual decision-making processes of consumers, incorporating heterogeneity in preferences and constraints, while the Bass model captures the cumulative adoption trend over time based on innovation and imitation effects. We apply this framework to analyze EV adoption rates in several Ecuadorian cities, using historical data on EV sales over multiple years. The simulation incorporates key factors that influence consumer decisions, including vehicle autonomy, battery capacity, maintenance costs, charging time, acquisition cost, and the number of EVs already in circulation. For this purpose, a parametric software tool was developed and implemented, capable of simulating and analyzing various EV adoption scenarios, considering multiple variables and environmental conditions. The results reveal that barriers to EV adoption are strongly tied to both individual vehicle characteristics and broader systemic constraints. These findings highlight the need for more effective policies and strategies that consider consumer behavior, economic factors, and infrastructure limitations. By addressing these barriers, countries like Ecuador can accelerate the transition to a more sustainable transportation system and contribute to significant reductions in carbon emissions within the automotive sector.
An urban consolidation centre (UCC) is a pivotal facility designed to improve the efficiency of urban distribution and mitigate the negative impact generated by urban logistics. However, many UCCs often fail to be self-sustaining because it is challenging to build sound partnerships amongst stakeholders with heterogeneous attributes and differing objectives. Existing literature identifies three types of barriers, namely: (1) financial and practical barriers, (2) social and cultural barriers, and (3) legal barriers, which would result in the stakeholders terminating their partnerships in the UCC project. In response, this study aims to identify critical factors in overcoming the above barriers, which will help contribute to the formation of successful partnerships in UCCs. We implement a comparative case study based on three UCCs, with three different partnerships, initiated in Sweden and China. We analyse the difficulties encountered by different stakeholders and the practical actions they took to deal with these difficulties. Based on the findings, this study identifies the barriers to stakeholder partnerships and concludes with several critical factors that can be applied to overcome the barriers.
Complexities in route planning for hazardous materials (HAZMAT) arise because of the fundamental need to optimize risk while considering the perspectives of regulators and transport carriers. The trade-off between risk and other criteria is skewed in the HAZMAT transportation context due to stringent regulatory requirements. Hence, risk is not weighted among other criteria but considered as a performance evaluator for other criteria. This reality changes the optimization goals and introduces complexities for the hazardous transportation network design problem (HNDP). It is evident that HNDPs solved in the literature often fail to penetrate actual practice. This article delivers a holistic review of the HNDP context from the literature, including risk regulatory setups and models, HNDP formulation challenges and variations, optimization approaches, and risk measures and constraints. The review reveals knowledge gaps that potentially deter the practicality of HNDP solutions. Specifically, more work is needed to develop optimization approaches that integrate non-linear cost functions and a mix of direct and indirect routing while following the user-optimal routing strategies. Data availability continues to limit such approaches. Addressing risk equity among communities and cost equity among transport carriers, requires tactical use of soft constraints and consideration of emergency response capacity. Moreover, network reliability may be critical for HAZMAT transportation and should be addressed to make HNDP solutions more practical. The review motivated the development of a novel regulator-carrier generic risk optimization framework that situates knowledge gaps within the complexities of the HNDP.
Transportation agencies are tasked with maintaining safety and mobility during adverse winter weather conditions. Agencies accomplish this objective by deploying snow and ice control strategies through their winter maintenance operations programs. These programs often entail use of decision support systems, maintenance vehicles for plowing and de-icing chemical applications, and, at times, road access restrictions. These agencies must assess their performance to identify strengths, weaknesses, opportunities for improvement, and threats to future success in maintaining optimal levels of service, mobility, and safety. This study assesses the Nebraska Department of Transportation’s winter maintenance operations performance during nine snowstorms during the 2023–24 winter season. Performance is assessed for each snowstorm for 20 road segments across Nebraska, with a focus on 12 segments along the Interstate 80 corridor. This analysis leveraged meteorological information (e.g., snowfall accumulation, wind conditions), traffic mobility insights (i.e., vehicle speeds), and winter maintenance operation activities (e.g., the amount of materials used, duration of operations) to understand and quantify the performance of winter maintenance operations. It was found that time, or duration, was the most meaningful performance metric. For example, the duration of snowfall or the duration of blowing snow conditions had stronger correlation with the duration of mobility disruptions or duration of maintenance operations than the amount of snow accumulation. The most important conclusion of this study is that a holistic consideration across multiple metrics offers the greatest benefit in the assessment of winter maintenance operations.
Despite its substantial contribution to road traffic collisions, injuries, fatalities, and associated economic costs, drowsy driving remains a persistent transportation safety and public health concern. Although there is substantial literature regarding sleepiness and fatigue in the context of driving, alongside policies, practices and safeguarding procedures, levels of drowsy driving appear to remain consistent, highlighting the need for a more integrated understanding of the factors influencing this behaviour. This narrative review synthesises interdisciplinary evidence from transportation, sleep science, psychology, occupational health, and public health. It examines the scale of drowsy driving, its contributing factors, and its emergence as a complex transportation behaviour shaped by interacting individual, social, and societal influences rather than solely by individual choice. Building on the reviewed evidence, a socio-ecological model of drowsy driving is proposed to integrate biological, behavioural, occupational, technological, and policy influences within a single conceptual framework. The review identifies key knowledge gaps and future research priorities and provides a framework to guide interdisciplinary transportation research, policy, and practice addressing drowsy driving and improving road safety.
Despite growing attention to urban travel carbon emissions, the differentiated drivers of commuting and non-commuting behaviors remain underexplored. Using Wuhan as a case, this study applies time-geographic theory and interpretable machine learning to examine nonlinear relationships shaping travel-related carbon emissions (TCE) across commuting and non-commuting groups. Results show commuting TCE is primarily influenced by socioeconomic factors such as gender and income, reflecting rigid constraints, while non-commuting TCE is more responsive to built environment features, especially POI density and transit accessibility. A shared low-emission window exists, with optimal POI density (100–400 items/km2) and intersection density (<15/km2). Interaction patterns indicate that built-environment associations vary across socioeconomic groups. The analysis also highlights Wuhan’s spatial duality: morphological polycentrism coexists with functional monocentrism, while non-commuting TCE is more closely associated with community-scale services. These findings suggest the need for differentiated, group-specific low-carbon strategies, contributing new evidence for integrating behavioral heterogeneity into transport and land use policies.
The paper investigates the driving behavior patterns of mainline human-driven vehicles during highway ramp-merging interactions. Using the exiD dataset, a dynamic merging window was defined in time and space to identify mainline vehicles with potential interactions with ramp vehicles. Using the neighboring-vehicle information for each ramp vehicle within the merging window, one-to-one merging events between mainline and ramp vehicles were extracted. To identify heterogeneous driving behavior patterns within the merging area, a Bayesian Gaussian Mixture Model was employed for cluster analysis of driving behaviors. Subsequently, the Hierarchical Bayesian Inverse Reinforcement Learning framework was adopted to infer global and cluster-specific bias weights that represent relative reward preferences across behavioral patterns. Afterward, using reinforcement learning, the driving behavior trajectories of human-driven vehicles in each cluster were reproduced in SUMO (Simulation of Urban MObility), providing a more comprehensive representation of lateral and longitudinal driving behavior. Finally, the exiD-trained policy was deployed in a high-fidelity SUMO merging scenario representing the Hungarian M1 highway. The results showed stable closed-loop execution and supported the cross-scenario portability of the framework. Overall, this study presents a configurable empirical framework for on-ramp merging analysis that connects traffic interaction representation, human driving behavior heterogeneity, data-driven preference inference, and closed-loop traffic simulation.
