Approccio proattivo alla sicurezza dell'infrastruttura stradale mediante valutazione delle condizioni operative del traffico e del contesto territoriale

The research project aims to develop an innovative model for proactive road safety assessment, a critical issue in the management of transport infrastructure. Traditionally, safety improvement strategies have relied on a reactive approach, focused on analysing past accidents, which has proven limited in preventing future incidents. To address these shortcomings, Directive 2019/1936/EU and recent technological advancements advocate for the adoption of predictive models capable of identifying latent risk factors and preventing accidents, thereby enhancing the overall safety of road networks. In this context, the project proposes an advanced methodology for proactive road safety analysis, based on observing operational traffic conditions, particularly actual operating speed, and the surrounding territorial context through a new indicator known as the Settlement Ratio (SR). The research integrates historical accident data, information gathered from probe vehicles (Historical Car Data), and spatial data related to the territory. The use of GIS tools and advanced computational software enables the mapping of both operational and territorial risks, identifying road sections with a higher concentration of incidents. This approach has been applied to selected rural roads managed by ANAS in Italy’s Veneto region, with the aim of providing targeted solutions to improve safety and optimize infrastructure management. The need for a proactive approach is particularly pressing for rural roads, where the number of accidents is lower compared to urban areas, but the mortality rate is significantly higher. The proposed methodology addresses this issue by considering both infrastructure characteristics and surrounding territorial features in an integrated manner. The originality of the project lies in the integration of diverse data sources and multidisciplinary approaches to offer a fresh perspective on road safety analysis. The most innovative aspect is the inclusion of the territorial context as a key factor in risk assessment. Taking settlement characteristics into account allows for tailoring infrastructure planning and management to the specific needs of local communities. Both urban and rural development directly influence traffic flows and require targeted interventions to ensure a safer and more efficient road network. Another distinctive feature of the project is the use of advanced analytical tools to better understand traffic dynamics and user behaviour, enabling the identification of latent risks along the road network. This approach facilitates timely and targeted interventions, improving resource allocation and reducing the likelihood of future accidents. Within the proposed methodology, the Settlement Ratio (SR) plays a pivotal role as an indicator measuring the level of anthropization in the area surrounding a road infrastructure. Unlike traditional tools, the SR provides a continuous representation of settlement conditions along the route, highlighting critical situations resulting from the interaction between infrastructure and territory. This indicator is complemented by the Settled Area (SA), which measures the proportion of built-up area within the study region. This approach provides an innovative tool to optimize road network management by enhancing its performance and adapting it to evolving territorial dynamics. An important future application of this methodology concerns road networks with limited or unavailable accident data. In such cases, a detailed analysis of infrastructural, territorial, and functional characteristics will enable the anticipation of the most common types of accidents, providing a valuable and proactive tool to support the management and planning of road infrastructure. Methodology The proposed methodology is structured into four main phases, each contributing to a holistic understanding of the operational conditions of the road network and the surrounding territorial context. Each phase has been designed to address specific analytical needs, ensuring a systematic and replicable process: 1. Road Network Geometry Analysis The first phase of the methodology involves reconstructing the road network geometry using georeferenced data extracted from the road graph. This process provides an accurate representation of the layout, essential for subsequent analyses. A MATLAB algorithm processes raw data to identify the planimetric elements of the infrastructure: straight sections are recognized as segments with constant azimuth, while circular curves are detected through smoothing techniques based on least squares fitting, which determine their radius and length. Based on the extracted planimetric data, the code generates a curvature diagram and calculates theoretical design speeds in compliance with the 2001 Italian Ministerial Decree. An integrated graph is then produced, overlaying the curvature and speed diagrams, enabling simultaneous analysis of the road layout and associated design speeds. This tool allows for evaluating the consistency between the road alignment and the intended operational conditions. The analysis goes beyond the road layout by including the identification of key infrastructural interferences, such as intersections, access points, bridges, and viaducts, which are critical elements influencing user behaviour and safety conditions. Correct identification of these interferences is essential to assess their impact on traffic dynamics and accident occurrence. The alignment consistency check is based on comparing geometric characteristics with theoretical design speeds. The study of curvature and speeds helps identify potential geometric issues that could compromise operational conditions and user safety. Finally, the integration between infrastructure and territorial context is analysed through the mapping of interferences to evaluate their effect on traffic conditions and the occurrence of accidents. 2. Settlement Analysis and Introduction of the Settlement Ratio (SR) One of the innovative elements of the proposed methodology is the introduction of the Settlement Ratio (SR), an indicator that quantifies the degree of anthropization along road infrastructures. The SR measures the ratio between built-up areas and the total plot surface, integrating building density and land use. This indicator provides a continuous representation of settlement conditions along the route and an assessment of their impact on traffic operational conditions. The SR is divided into four macro-categories: residential, community, infrastructural, and agricultural, offering a detailed view of the functions of buildings and the interactions between infrastructure and territory. The methodology involves the use of georeferenced data processed through GIS software like QGIS, which integrates information on buildings, land cover, and the road graph, and computational tools like MATLAB to implement and process the data. The methodology combines spatial analyses and functional data to provide a detailed and accurate representation of the territorial context. The Settlement Ratio (SR) allows the evaluation of both the number of built-up areas in a specific region and the functions of the buildings, offering a comprehensive measure of the degree of anthropization. This index synthesizes building density and land use characteristics, making it essential for understanding the interactions between the built environment and road infrastructure. Data collection and processing are conducted using QGIS software, which allows for the integration of georeferenced data on buildings, land cover, and the road graph, enabling location-based analyses. The spatial analysis utilizes moving investigation windows, represented by circles with a 500-meter radius centred on road network vertices. This approach systematically identifies all buildings and plots within the area surrounding the analysed road section, ensuring a precise and continuous assessment of settlement conditions along the entire infrastructure. The moving windows ensure a progressive and continuous representation of settlement conditions, placing the road infrastructure at the centre of territorial analysis. The collected data is then subjected to a verification and validation process to ensure quality and reliability for subsequent analyses. The identified buildings are aggregated into a single dataset, and territorial plots are incorporated into their respective new dataset. The data validation includes geometry checks to detect and eliminate errors such as self-intersections and duplicates, as well as the aggregation of information to optimize computational operations and reduce redundancies. The SR calculation is performed in MATLAB and is structured into two main phases. The first phase involves a preliminary calculation of the ratio between the built-up area of a plot and its total surface. The second phase consists of weighting the initial value using a multiplicative coefficient that accounts for the plot size relative to the investigation window. This step ensures that the SR accurately reflects plot characteristics, preventing smaller plots from disproportionately influencing the final results. The final SR value ranges from 0 to 1, where higher values indicate a greater degree of anthropization, signalling a significant presence of buildings and settlements in the analysed territory. A key aspect of the methodology is the classification of land uses. To simplify and optimize the analysis, land uses are grouped into four main macro-categories: • Agricultural: Including areas designated for agricultural activities and natural spaces. • Infrastructural: Covering areas occupied by road, rail, and service infrastructures. • Residential: Refers to areas designated for housing buildings. • Community: Encompassing public, commercial, and industrial structures. This classification allows the specific contribution of each category to the overall SR value to be quantified, providing a detailed view of the different types of settlements along the road route. The analysis results are graphically represented along the examined road sections. The final graph shows the trend of the SR for each macro-category and the percentage of Settled Area (SA), which represents the portion of built-up territory within the investigation windows. This visualization makes it easy to identify areas with high building density and understand how urbanization impacts road infrastructure operational conditions, providing essential support for road network planning and management. The second phase of the proposed methodology enables a detailed analysis of settlement characteristics along specific road sections, using graphs that illustrate the Settlement Ratio (SR) and the Settled Area (SA). These analytical tools help understand the distribution and density of buildings along roadways, highlighting how roads cross areas with varying building densities, alternating between urbanized zones and rural contexts. Peaks in SA are significant near urban centres, where high SR values for community and residential categories are prevalent, while they decrease in more rural areas. However, even in extra-urban zones, the indices are not negligible, indicating that road infrastructures are influenced by anthropization even in the absence of urban centres. The conducted analysis provides a refined and validated dataset essential for rigorous spatial analysis, useful both for understanding settlement distribution along roads and for improving territorial planning. The proposed methodology identifies areas where urbanization significantly impacts road infrastructures, providing operational support for territorial management. Understanding the distribution and intensity of settlements along roadways is crucial to ensure that residential and community areas are adequately served by infrastructure and services, promoting balanced and sustainable territorial development. This methodology also allows for analysing territorial dynamics with a proactive approach, identifying critical points where high surrounding settlement levels can influence operational conditions and road safety. Areas with high SR values tend to present more complex traffic dynamics, potentially increasing accident risk. In this sense, integrating accident history analysis can provide valuable information for developing targeted road infrastructure management strategies. The methodological approach described also serves as a useful tool for future planning. By identifying rapidly expanding areas or zones at risk of congestion, it is possible to anticipate infrastructure needs and plan timely interventions to improve road capacity and reduce the likelihood of accidents. Risk prediction and management are essential to ensure sustainable infrastructure development in response to population growth. Analysing SR variations is a key element for improving safety and efficiency in long-term road planning. 3. Operational Speed Analysis In the third phase, operational speeds along the road network are analysed. This process relies on historical data collected through probe vehicles, processed using MATLAB to generate a continuous speed profile that reflects actual user behaviour. The analysis identifies discrepancies between regulatory design speeds and the speeds drivers adopt, pinpointing critical sections where road geometry or functionality leads to deviations from expected driving behaviour. To assess mobility and the operational efficiency of road infrastructure, Historical Car Data (HCD) provided by Infoblu, a company within the Atlantia Group, were utilized. The dataset, gathered from roads in the Veneto region, includes approximately 11.5 million speed measurements recorded during three specific periods: August 2018, February 2019, and May 2019. The goal of the analysis was to evaluate traffic conditions by constructing an operational speed profile, defined as the 85th percentile (V85) of speeds practised by users under ideal conditions. The reconstruction of operational speeds began with data management using MySQL Workbench. The map-matching technique was applied to project GPS coordinates onto road graphs. The data were then divided by travel direction (South-North and North-South) and filtered to create a precise speed profile. This filtering process excluded data influenced by adverse weather or traffic conditions, ensuring a reliable assessment of driving behaviour along the monitored sections. The analysis showed that the studied road sections operated under free-flow traffic conditions, without significant congestion. The data were further segmented into two main periods: daytime and nighttime. This segmentation enabled the creation of a geolocated dataset along the road’s curvilinear reference line, useful for developing a continuous speed profile using the smoothing cubic spline technique. Comparing operational speed profiles between daytime and nighttime periods revealed notable differences in driver behaviour, influenced by traffic density, road geometry, and risk perception. During nighttime hours, operational speeds tended to be higher than those recorded during the daytime, primarily due to reduced traffic congestion and a diminished perception of risk, leading drivers to adopt less cautious behaviour. However, in sections with restrictive geometry or specific territorial conditions, speed reductions were observed even at night to maintain safety. Conversely, daytime operational speeds were generally lower than at night. This difference was particularly evident in extra-urban industrial areas, where heavy traffic and work-related activities negatively affected speeds, resulting in values below theoretical design speeds. These discrepancies were most pronounced during weekday time slots, when industrial traffic intensity was at its peak, causing noticeable slowdowns in vehicle flow. The comparison between operational and theoretical speeds showed frequent deviations from expected values. Graphical representations helped identify critical sections of the road network where real operational conditions did not align with design assumptions. These deviations were attributed to a combination of geometric, environmental, and behavioural factors that influence driving behaviour and may increase accident risk. The speed profile analysis confirmed that variations between operational and theoretical speeds were driven by infrastructure characteristics, environmental conditions, and visibility factors. These discrepancies indicate that drivers adjust their behaviour to match real-world road conditions, modifying their speeds in response to contextual cues. The insights gained from this analysis are essential for developing targeted traffic management strategies tailored to daytime and nighttime conditions. These strategies aim to enhance road safety and improve the overall efficiency of infrastructure operations. Monitoring critical sections helps identify areas where inconsistencies between user expectations and road characteristics may pose safety risks. In such cases, targeted corrective measures are necessary to ensure safer conditions, reduce accident likelihood, and improve the overall driving experience. 4. Accident Analysis through Kernel Density Estimation (KDE) Analyzing accident data is essential for identifying critical points along the road network and understanding the factors influencing infrastructure safety. To obtain a continuous spatial representation of risk, the proposed methodology employs Kernel Density Estimation (KDE), a non-parametric statistical tool that estimates the probability density of accidents along the road network. This approach overcomes the limitations of point-based analyses by focusing not only on the locations where accidents have occurred but also on surrounding areas that may be at risk. The application of the KDE function represents accident data as a continuous curve rather than a collection of isolated points. This representation extends the safety analysis to the entire road network, maintaining a non-zero accident density value even in sections where no accidents have been recorded but which may still be exposed to potential risks. This approach reflects the idea that accidents are not isolated events but rather the culmination of a risk process that begins upstream of the critical point along the infrastructure. By qualitatively evaluating the KDE curve, it is possible to identify accident peaks corresponding to high concentrations of incidents and analyze the extent of the “bell shapes,” which highlight risk influences both upstream and downstream of the critical point. This analysis identifies areas requiring priority interventions, providing a solid foundation for planning maintenance and infrastructure improvements. Although the proposed methodology is proactive and based on future risk predictions, the validation of the research project required the use of historical accident data. These data allowed for the analysis of recurring events and the evaluation of the impact of infrastructural characteristics on risk factors along the road network. Historical data analysis revealed a strong pattern between accident types and the geometric, environmental, and traffic conditions associated with each road section. One notable finding was the homogeneity of accident profiles observed in sections with similar characteristics. This made it possible to associate specific accident types with each homogeneous road section, identifying recurring risk patterns. The analysis used georeferenced accident data provided by ANAS S.p.A., related to incidents recorded in the Veneto region between 2015 and 2019. The data were imported into QGIS and positioned according to the WGS84 reference system, then converted to the Gauss-Boaga system to ensure compatibility with the road graph used in subsequent MATLAB analyses. During the map-matching process, each accident event was associated with its corresponding position along the road network, and essential information (geolocation, nature, and severity of the incident) was filtered and stored in the accidents.mat file. The KDE processing in MATLAB generated a continuous profile of accident hotspots across the road network. This method identifies spatially distributed critical points, including sections surrounding accident points as potentially at risk. This approach surpasses traditional analyses, which focus solely on precise accident coordinates, offering a more comprehensive and dynamic view of criticalities along the road network. Building a continuous profile of accident density enables the identification of distributed risk areas along the road network, facilitating the planning of proactive interventions. The KDE curve broadens the analysis to include potentially risky sections, even in the absence of concentrated accident clusters, emphasizing the role of infrastructural, environmental, and behavioural factors. In particular, directly comparing the accident density curve with operational speed and settlement profiles identifies road sections with the highest exposure to accident risk. This approach highlights critical sections where operational conditions or territorial configurations influence user behaviour, increasing the likelihood of incidents. The proactive approach correlates accident density data with geometric, infrastructural, and behavioural factors, providing a comprehensive understanding of risk dynamics across the road network. This analysis is fundamental for planning and programming targeted interventions aimed at eliminating or mitigating risk factors present along the road. In conclusion, the proposed methodology enables a dynamic and precise approach to managing critical points along road infrastructures, fostering a planning process grounded in risk predictions and the integration of operational, settlement, and geometric data. This approach ensures greater effectiveness of interventions, improving road safety in a sustainable and targeted manner. Application of the Methodology The project was initially tested on a set of extra-urban roads managed by ANAS in the Veneto region, with a primary focus on the Strada Statale 309 "Romea," one of the busiest and most critical arteries in the region. Subsequently, the methodology was applied to other strategic roads to verify its flexibility and adaptability to different territorial contexts. The sample of roads included the following routes: SS14, SS47, SS50, and SS309. This validation phase confirmed the robustness of the methodology and its ability to adapt to various infrastructural configurations. During the analysis, cadastral and land cover data were obtained through the Veneto Region Geoportal. These data were processed using GIS software such as QGIS, enabling advanced visualization and manipulation of spatial information. Additionally, the automation of computational processes through MATLAB allowed for efficient data management and rapid implementation of operations. Automation is a key element of the methodology, as it reduces analysis time and enhances the precision of results. Results and Analysis of Findings The analysis of the road network revealed a significant pattern between accident types and the infrastructural and territorial characteristics of each section. The results indicate that the most frequent types of accidents along the analyzed routes are rear-end collisions and side-impact crashes, with dynamics varying based on road geometry, settlement density, and traffic conditions. In extra-urban sections with straight configurations and minimal settlement presence, rear-end collisions are predominant. This trend can be attributed to high operating speeds and drivers' failure to maintain safe following distances, exacerbated by the monotony of the route and limited visual stimuli. For example, on the SS309, rear-end collisions account for 56.8% of accidents in some straight sections, while on the SS50, this figure rises to 71.4%. These findings highlight the need for targeted interventions, including improved road signage and awareness campaigns promoting compliance with safe following distances. In urban sections or those with numerous private accesses and intersections, such as the SS47 and SS14, side-impact crashes are more prevalent. In these areas, fragmented vehicle flow, reduced visibility, and increased conflict points due to frequent entries from side accesses contribute to higher accident rates. Side-impact crashes represent up to 55% of accidents in certain urban sections, followed by rear-end collisions, which account for approximately 25%. Managing access and improving visibility are essential measures to reduce accident rates in these cases. In sections with more complex road layouts, such as sharp curves or winding stretches, run-off-road accidents are particularly frequent. This is especially evident along the SS14 and parts of the SS309, where heavy traffic and adverse weather conditions increase the risk of vehicles leaving the roadway. In some of these sections, run-off-road accidents account for 30.8% of incidents, underscoring the importance of improving pavement quality and horizontal signage to minimize the risk of vehicle loss of control. The analysis demonstrated that settlement density significantly impacts road safety. In high-density areas, the presence of buildings, private accesses, and intersections increases conflict points, favouring side-impact crashes. Conversely, in low-density areas, the absence of settlements and straight road configurations leads to a higher occurrence of rear-end collisions, often caused by high speeds and driver distraction. In summary, a combination of infrastructural and behavioural factors significantly affects risk patterns along the road network. In extra-urban straight sections, rear-end collisions pose the highest risk, whereas, in urban areas or sections with numerous side accesses, side-impact crashes are more common. Additionally, in sections with restrictive curves or complex geometries, run-off-road accidents are frequent, particularly during heavy traffic or adverse weather conditions. Based on these findings, targeted interventions are recommended to enhance road safety. Suggested actions include optimizing road geometries, implementing more effective signage, and organizing awareness campaigns to promote safe driving behaviours. Furthermore, the adoption of advanced technological systems for real-time traffic condition monitoring could support more efficient infrastructure management, thereby reducing accident rates. Conclusions The primary outcome of this project is the development of an innovative methodology for road safety analysis. It offers an advanced approach to evaluate the influence of spatial context on road infrastructures. The proposed methodology integrates complex data related to infrastructural, settlement, and operational conditions, providing a holistic view of the road network and enhancing the understanding of multifactorial interactions between infrastructure elements and the surrounding environment. Its proactive nature sets it apart, enabling the identification of recurring accident types, the rapid detection of critical sections, and the planning of targeted interventions to improve road safety, even in the absence of updated accident data. This methodological approach marks a significant improvement over traditional ex-post analyses, promoting more effective and sustainable road infrastructure management. A key strength of the project is the use of software tools like QGIS and MATLAB, which enabled the automated integration of complex parameters, including operational speeds, land use, and accident dynamics. This process allowed for the creation of a detailed and continuous overview of the entire road network, highlighting the relationship between territorial factors and accident distribution. Another noteworthy contribution of the project is the development of specific indicators, such as the Settlement Ratio (SR) and Settled Area (SA), which provide a detailed mapping of the degree of anthropization and its impact on road safety. These indicators facilitate an in-depth analysis of the interactions between urban context and road network operational conditions, offering a solid foundation for understanding how human development levels affect accident risk. A key concept introduced by the project is that of Homogeneous Sections, which divide the road network into segments characterized by similar user behaviour and settlement conditions. This segmentation enables a more precise analysis of driving dynamics, identifying points where geometric or environmental conditions may negatively impact safety. The proposed methodology underscores the need for future studies to establish a statistical correlation between qualitative and quantitative factors characterizing the infrastructure and its surrounding environment. The objective is to better understand the multifactorial interactions between road characteristics, operational conditions, and accident distribution, identifying the factors that contribute most to accident risk and proposing targeted interventions based on the nature of the most frequent events. Integrating this methodology with advanced monitoring systems, such as IoT sensors and drones, represents a promising development that would enable the collection of real-time traffic data, enhancing the model's predictive capacity and facilitating proactive road safety management. In summary, the project proposes an integrated and multidisciplinary approach that combines infrastructural, operational, and territorial analyses to accurately identify critical road sections and plan targeted corrective measures. The model’s strength is its capacity to deliver precise and customized insights for risk management and mitigation, making it applicable to various contexts and promoting dynamic and sustainable road network management. This approach marks a significant advancement over traditional ex-post analyses by preventing critical issues and improving road safety, ultimately reducing accident rates and ensuring efficient infrastructure management in a complex and ever-evolving mobility landscape.