Periodic Reporting for period 1 - SOTERIA (Systematic and orchestrated deployment of safety solutions in complex urban environments forageing and vulnerable societies)
Reporting period: 2022-11-01 to 2024-04-30
Although the European roads are the safest in the world, 20 634 people lost their lives in 2022 with almost 70% of fatalities in urban areas being vulnerable road users (VRUs). The EC has adopted the Road Safety Policy Framework 2021 – 2030, as part of its efforts to halve the number of fatalities by 2030. However, progress in reducing fatality rates remains below the interim target of 50%, a milestone on the way to achieve close-to fatalities by 2050, the so-called "Vision Zero". At the same time, more and more people opt for active forms of mobility for reasons of personal health and fitness or for environmental considerations. Measures and policies promoting active mobility are considered a game changer in reducing CO2 emissions and congestion in urban areas. But safety needs to be systematically taken into consideration. Improved road infrastructure and lower speeds, for example, have the capacity to reduce the impact of collisions that may occur. SOTERIA aims to accelerate the attainment of “Vision Zero” goals by providing a framework of models, services and tools that enable data-driven urban safety intelligence. Project partners are working towards designing solutions that promote safe and green travelling for VRUs, foster integration of electric micro-mobility in urban environments and enable inclusive transport for all road users. Their aspiration is to achieve road safety equality for the ever-changing urban mobility scene where disruptive services and complex needs necessitate innovative solutions. Project partner’s mission is to uncover unexplored behaviours of VRUs and engage Living Lab communities in the co-creation of urban safety solutions. Simulation models for the safety of public spaces and accident analysis as well as data-driven routing applications leveraging explainable Artificial Intelligence (XAI) are developed. A platform of interconnected services is designed to determine risk-free routes, improve the visibility of VRUs to motorised vehicles, inform VRUs of potential hazards and nudge them towards safer behaviours.
During the first 18 months, SOTERIA partners focused on the design of road safety solutions starting with the setup of Living Labs (LLs) in Germany, Greece, Spain and the United Kingdom. The SOTERIA LLs offer a contextual framework for engaging communities of VRUs in the acquisition of widely accepted solution specifications. Surveys are used to investigate stakeholder views towards data-driven technological solutions. Based on the data collected, 7 personas were identified describing user types that can benefit from the SOTERIA solutions. A use case was defined for each persona, covering different functionalities of 16 solutions. This work provided the foundations for the development of an integrated platform composed of advanced accident analysis, safe travelling services, micro-vehicle sensor kits and web/mobile applications for the end users. Close to 50 data sources were identified and integrated in the Safe Mobility Data Space that cover transport supply and demand, cartographic resources, driving behaviour and accident data. In parallel, the first releases of the accident data processing and analysis module and the labelled graph module were completed, along with a data visualisation module, three key components of the SOTERIA accident modelling and simulation suite. Initial versions of accident prediction models were implemented. The first, combines graph-based neural networks with other techniques to predict accident frequency and severity at section and intersection level by considering influencing factors and road conditions. The second, utilises generalised linear models to estimate the risk and number of accidents. The alpha versions of travel demand estimation and XAI modules were also released. The XAI module enhances the transparency, fairness, and reliability of accident predictions, while the travel demand estimation module reconstructs mobility patterns, fusing mobile network data, connected vehicle data, shared mobility data, traffic counts, etc., to determine travel routes. Two sensor kits were developed to allow the collection of data from micro-vehicles: one to collect environmental, motion-related, positioning and proximity data, and a kit collecting environmental data upgraded with the addition of motion-capturing sensors. These kits will provide complementary data to a dashcam capturing near-misses. The data collected will be used as input for the speed advisory module designed to provide real-time warnings to VRUs. Furthermore, a routing engine that advises VRUs of safe and environmentally friendly paths was developed. The first version of a mobile application with an integrated nudge engine was implemented to allow access to the aforementioned services with an overall goal the enhancement of VRUs’ safety in their day-to-day travel. Lastly, comprehensive guidelines for the pilot demonstrations and assessments have been elaborated as well as a plan to be pursued for the exploitation of the results in the later phases of SOTERIA and beyond its lifetime.
SOTERIA aims beyond the state of the art in various scientific domains as follows: •accident analysis and prediction, by the development of models using novel Deep Learning (DL) architectures including graph neural networks, convolutional neural networks, multi-view and multi-task neural networks and others; •XAI for accident prediction, by the investigation of XAI techniques to extract information from the DL models regarding factors involved in road safety and relations among them; •route engines for safe travelling, by the design of models and safety performance functions that allow the assessment of safe route alternatives, instead of merely identifying accident-conspicuous road network sections;•active safety systems for micro-vehicles, by developing sensor kits and proactive awareness models that assist riders of micro-vehicles to mitigate risks that may lead to accidents;•situational awareness services for VRUs, by the provision of a connected micromobility users’ platform that broadcasts location and behavioural attributes, customisable through context-aware mechanisms and supported by infrastructure-based monitoring for advanced perceptive functionality. Further enhancements to the functional versions of the above innovations are planned for the second half of the project.