D3.1 New and enhanced models/simulation environments and user support materials (Beta edition) - PUBLIC FACING VERSION

This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) z New and enhanced models/simulation environments and user support materials (Beta ed.) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Document Control Page Deliverable number 3.1 Deliverable title New and enhanced models/simulation environments and user support materials (Beta ed.) Deliverable version 1.0 Work Package number 3 Work Package Title Model development and enhancement for VRU and urban safety Due date of delivery 30/04/2024 Actual date of delivery 30/04/2024 Dissemination level SEN - Sensitive Type Other Editor(s) James Bradford, Monica Olyslagers and Shanna Lucchesi Contributor(s) Marcel Sala, and Mark Brackstone (AIMSUN), James Bradford, Monica Olyslagers and Shanna Lucchesi (iRAP), Arunava Putatunda, and Santhanakrishnan Narayanan (TUM), Maria Oikonomou and Stella Roussou (NTUA), Amir Pooyan Afghari, Amna Chaudhury and Amir Hossein Kalantari (TUD), Sam Chapman (The Floow) Reviewer(s) Marko Sevrovic (EIRA) and Apostolos Ziakopoulos (NTUA). Project name Predictive Approaches for Safer Urban Environments Project Acronym PHOEBE Project starting date 01/11/2022 Project duration 45 months Rights PHOEBE Consortium 2 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Deliverable executive summary 4 Deliverable 3.1 is an intermediary deliverable for activities 3.1, 3.2, 3.3 and 3.4. It combines all the model development and enhancements built up individually during the referred activities. D3.1 presents the strategies of development of the behavioural changes, mode shift and induced demand and it’s protocolos for represantation in the simulation scenarios (though the integration protocols), the enhancements for existing road safety assessment models and the enhancements and calibration of traffic simulation environments. The content will be posterior used in technical documentation and user support materials. Part of this content is already available and presented in the deliverable. The content is considered sensitive since the simulation environment is copyright protected. It is the first of WP3. It captures the preparation and development of the enhanced models, simulation environments and user support materials ready for testing in the PHOEBE use cases. Specifically it includes the preparation of several aspects for integration into the AIMSUN traffic simulation environment, including: •The mode shift and induced demand models to support dynamic outputs of the PHOEBE framework (i.e. results that reflect changing traffic flow, speed etc.) •The human behavioural models •Road safety assessment model enhancements, including the aligning the data segmentation with the links and nodes used in Aimsun, and how the dynamic risk ratings (based on variable speed and flows) can be produced; and •Upgrades to the Aimsun Next Simulation Environment. •Integration procedures and strategies to the representations of the behavioural changes and induced demand in the simulation scenarios, as well as the combination Deliverable 3.1 aims to: •Combine all the model development and enhancements that have been undertaken both individually and concurrently throughout the phases of WP3: (i) preparation for development, and (ii) actual development •Outline and document project progress and activities to be conducted in the next WP3 phase, and •Detail the preparatory measures taken to validate the PHOEBE Framework's efficacy within designated use cases. •Present the techincial documentation and user support materials available at this intermediary stage. The preparatory work and enhancements discussed in this deliverable are based on the theoretical principles and methodological approaches of PHOEBE framework captured in Deliverable 1.2. It is important to note that reading Deliverable 1.2 is essential to understand the developments presented in Deliverable 3.1. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Project Executive Summary The EU-funded ‘Predictive Approaches for Safer Urban Environment’ (PHOEBE) project aims to develop an integrated, dynamic human-centred predictive safety assessment framework in urban areas. This will be achieved by bringing together the interdisciplinary power of traffic simulation, road safety assessment, human behaviour, mode shift and induced demand modelling and new and emerging mobility data. Focused on vulnerable road users’ safety, the 3.5-year-long PHOEBE project will draw inspiration from real-world scenarios in the three pilot cities of Athens (GR), Valencia (ES) and West Midlands (UK). Testing activities will be performed across the use cases to simulate and forecast the impact of changes on safety in different scenarios of disruptions or transitions across urban transport networks. Predicting and visualising the safety and socioeconomic outcomes of new forms of transport, new technologies, or regulatory and behavioural changes from the individual (micro) level up to the network-wide (macro) level will also be a significant game-changer for urban stakeholders. The results of PHOEBE can be used as a blueprint by other European cities to develop their knowledge products, such as socioeconomic analysis model, urban road safety assessment, human behaviour and choice modelling. PHOEBE pilot cities List of participating cities: •Athens (Greece) •Valencia (Spain) •West Midlands (United Kingdom) Social links: https://twitter.com/Project_PHOEBE https://www.linkedin.com/company/phoebe-project/ https://www.youtube.com/@phoebeproject For further information please visit WWW.PHOEBE-PROJECT.EU 5 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Project partners Organisation Country Abbreviation EVROPSKI INSTITUT ZA OCENJEVANJE CEST -EURORAP SI EIRA ETHNICON METSOVION POLYTECHNION EL NTUA TECHNISCHE UNIVERSITEIT DELFT NL TUD TECHNISCHE UNIVERSITAET MUENCHEN DE TUM AIMSUN SLU ES AIM POLIS AISBL BE POLIS FACTUAL CONSULTING SL ES FC UNIVERSITAT POLITECNICA DE VALENCIA ES UPV OSEVEN SINGLE MEMBER PRIVATE COMPANY EL O7 THE FLOOW LIMITED UK FLOOW INTERNATIONAL ROAD ASSESSMENT PROGRAMME UK iRAP 6 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) List of abbreviations and acronyms Acronym Meaning AADT Average Annual Daily Traffic AI Artificial intelligence API Application Programme Interface ATH Athens CBA Cost Benefit Analyses DX.X Deliverable X.X FSIs Fatal and Severe Injuries GIS Geographical Information System KPIs Key Performance Indicator ML Machine learning NCBs Non -Compliant Behaviours OD Origin Destination P Probability 7 Acronym Meaning PuT Public Transport RDS Real data sets SEA Socioenomic Analysis SP Stated preference (survey) SR Star Ratings TAZ Traffic Analysis Zone VAL Valencia VKT Vehicle -kilometers traveled VRU Vulnerable road users (i.e. pedestrians, bicyclists, motoryclists and other PTW and micro -mobility user groups) WM West Midlands This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Table of Contents 1. Phoebe Overview 2. Model specifications 3. Integration protocols 4. Models validation, calibration and testing 5. User documentation 6. Use cases dedicated models 7. Final remarks 8 Deliverable 3.1 is considered sensitive, and part of its content was excluded from the public version. Please contact the project if you want to learn more about any particular aspect presented in this document. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Progress beyond the state of the art 9 Deliverable 3.1 advances the current state of the art for incorporating human behaviour and choice models to support safety simulation and analysis. This deliverable is the initial attempt in the PHOEBE project to test its advances of the SoA in the following areas: •Integrating human behaviour and modal shift models into traffic simulation for safety analysis purposes: The aim of this work is to capture of the impacts of transport disruptions or transitions, by addressing behavioural adaptation, travel behaviour change and infrastructure provision. •Improving traffic simulation models for Vulnerable Road User (VRU) safety: The aim of this work is to establish a methodological process for incorporating non-motorised/light transport modes in traffic simulation to predict road safety outcomes for these road user groups. •Enhancement of road safety assessments for urban road networks: The aim of this work is to enhance existing road safety assessment models, namely iRAP’s Star Rating and FSI estimations with a focus on urban environments and CycleRAP for micro-mobility simulations, and associated user tools with the capability for dynamic modelling of road user risk and scenario testing. •Socioeconomic impact assessment: The aim of this work is to devise a harmonised methodology for the socio-economic impact assessment of safety measures at three levels (health and safety, environmental and economic) through network-level extrapolation road safety evaluation mode shift and induced demand modelling components of the PHOEBE framework to allow informed decision making by all stakeholders. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 1) PHOEBE Project Overview and Deliverable Objectives This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand model PHOEBE developments 17 Demand models relate to how many people want to access aparticular mode of transport. They considers the choices people will make when presented with two or more modes to choose from (‘mode choice), and whether somebody will make a trip at all based on the available modes (‘induced/suppressed demand’). The demand model encompasses all necessary estimations to achieve two primary objectives: (i) informing simulation models about modal share and modal shift, (ii) estimating induced demand for interventions, and (iii) generating important input values, e.g., Value of Preventing Fatalities, Value of Time etc.for the socio-economic analysis. Developments presented in Deliverable 3.1: •Modelling hypotheses •Model types and their formulation •Endogenous and exogenous variables •Model parameters •Examples of models and datasets •Planned SP survey Developments hypothesis: •Socio-economic and socio-demographic factors affect mode choice, modal shift and induced demand. •Characteristics (quantifiable) associated with modes of transport, e.g., cost, time etc. affect mode choice, modal shift and induced demand. •Characteristics (non-quantifiable) associated with modes of transport, e.g., attitudes, comfort etc. affect mode choice, modal shift and induced demand. •Mixing stated preference and revealed preference data may provide better forecast of mode choice, modal shift and induced demand. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand models Overview 18 The demand modelling component of the PHOEBE framework consists of two items: •Mode choice modelling •Induced demand modelling Mode choice modelling:This modelling component will deliver the usage probabilities (modal share) of different modes and the shifts occurring in these probabilities (modal shifts) as a function of the exogenous variables, e.g., travel cost, travel times, risks associated, etc. Induced demand modelling:This modelling component will deliver the number of excess or fewer trips made by the travellers as a function of the exogenous variables, e.g., travel cost, travel times, risks associated, etc. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand models Overview 19 Continue to other components of PHOEBE development This flowchart is a blown-up diagram of the Demand Models (Step 1) shown in Figure 2-1 of Deliverable D1.2. In Step 1 (Fig 2-1, D1.2), the demand models will be estimated calibrated and validated. In the later stages (Step 2), these models will be implemented with the changes. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand models Model types and their formulations 20 Discrete choice models (Random Utility Model (RUM)) Our investigation starts with Multinomial Logit Model (MNL). •Endogenous variable: 𝒚𝒏𝒊 (𝒃𝒊𝒏𝒂𝒓𝒚,𝒊.𝒆.,𝒀𝒆𝒔𝒐𝒓𝑵𝒐) à For mode choice (Particular mode used àYes/No) and induced demand (New trips made à Yes/No) •Exogenous variable (quantifiable characteristics): 𝒙𝒏𝒊𝟏,𝒙𝒏𝒊𝟐,… •Coefficients: 𝜶𝒊𝟎,𝜶𝒊𝟏,… 𝒚𝒏𝒊 =41,𝑖𝑓𝑉&' +𝜀&' >𝑉&' +𝜀&' ∀𝑗 0,𝑂𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 ,𝑤ℎ𝑒𝑟𝑒:𝑉𝑖𝑠𝑑𝑒𝑡𝑒𝑟𝑚𝑖𝑛𝑖𝑠𝑡𝑖𝑐𝑎𝑛𝑑𝜀𝑖𝑠𝑟𝑎𝑛𝑑𝑜𝑚𝑢𝑡𝑖𝑙𝑖𝑡𝑦. 𝑉&' =𝛼'( +𝛼').𝑥&') +𝛼'*.𝑥&'* +⋯ 𝜀&'~𝐺𝑢𝑚𝑏𝑙𝑒(0,𝜇) 𝑉𝑎𝑟 𝜀&' =𝜋* 6𝜇*,𝑤ℎ𝑒𝑟𝑒𝜇𝑖𝑠𝑡ℎ𝑒𝑠𝑐𝑎𝑙𝑒𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟𝑠 Core assumptions: •Independence of Irrelevant Alternatives (IIA): The relative probability of someone choosing between two options is independent of other alternatives in the choice set. •𝜺𝒏𝒊’s are Gumble distributed (Extreme Value Type 1) •Variance is inversely related to scale parameter 𝝁 •𝜺𝒏𝒊’s are independent and identically distributed (IID) •No correlations across alternatives and observations •Extent of noise across alternatives and observations is the same This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 21 Binary probabilistic choice (Yes or No) from stochastic model Probability density of random utility of MNL Stochastic / random utility models •Describe the preferences and choices in a probabilistic way •Instead of predicting the choice that a decision maker will make with certainty, they predict the probability of choosing each alternative (sigmoid) Demand models Model types and their formulations Stochastic utility (ε) •ε does not have a specific value, but follows a distribution •The type of model depends upon this distribution, e.g., the ε is normally distributed in a Probit model. The model specification depends on the data quality of data gethered from SP surveys. MNL is the starting point of the investigation and the model type and complexity will be gradually increased for accommodating more complex interactions of the exogenous variables. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 22 The following figure elaborates the linkage of the inputs and outputs of the demand models with other components of PHOEBE’s framework. The demand models connect with road safety assessment models and AIMSUN microsimulation models through its inputs and outputs. Demand models Linkage (input/output) with other components of PHOEBE’s framework Safety is incorporated into the demand model in two components: (1) Road safety assessment model results as a proxy of perceived infrastructure risk (2) Risk profile of modes investigate in the Sp scenarios 1 2 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 23 Random utility model (MNL) Factor analysis Measurement equation Demand models Model types and their formulations We will investigate machine-learning techniques for non-parametric random utility 𝜺 -advancing SOA of traditional mode shift models. The availability can only be testified when mode survey data is available for starting model development. For non-quantifiable exogeneous variables, (e.g., perceived risk, attitudes, comfort etc.) we employ latent variables and hybrid choice models. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 24 Demand models Model types and their formulations The 𝑋'are the explanatory variables for the demand models while 𝐼are the indicators for the latent variables. Hybrid choice and latent variable model example Choice model This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand models Endogenous and exogenous variables and model parameters •Endogenous variable: Choice(s) (Yes or No) •Exogenous variables: •Socio-demographic variables: Age, gender, education, marital status, household, etc. •Socio-economic variables: Income, spending, occupation, employment, etc. •Travel behaviour variables: Travel destination, purpose, modes used, choices, etc. •Quantifiable variables: Travel cost, travel time, risk associated, etc. •Non-quantifiable variables: Attitudes, comfort, weather conditions, etc. •Sources: Stated preference surveys, Revealed preference data. 25 The model parameters of importance are the coefficients of the exogenous variables of the deterministic utilities (𝜶𝒊𝟎,𝜶𝒊𝟏,…) 𝑽𝒏𝒊 =𝜶𝒊𝟎 +𝜶𝒊𝟏.𝒙𝒏𝒊𝟏 +𝜶𝒊𝟐.𝒙𝒏𝒊𝟐 +⋯ The parameters of the models will be estimated, calibrated followed by the validation of the models. These parameters will be further used to calculate values like: •Willingness to pay (for socio-economic analyses) •Value of Time (for socio-economic analyses) •Maximum acceptable risk, and •Willingness to wait. This presents the exogenous and endogenous variables with the utility of the estimated demand models’ parameters. Application in the PHOEBE project This information serves as a valuable guide for future applications of PHEOBE, delineating the requisite model parameters essential for accurate utility estimation. It is imperative to underscore that acquiring this information necessitates either leveraging existing mobility surveys/data or strategizing new surveys to gather both stated and revealed preference data. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand models Data requirements 26 Data requirements for demand models An array of data is required to estimate, calibrate and validate travel demand models (including the induced demand models) for the current state recreation future forecast. Deliverable 1.2 includes Table 2-1, which presents a comprehensive list of data requirements and their sources. In the case of PHEOBE, existing data sources fell short of meeting the project's demand model requirements, necessitating the planning and execution of new surveys. More on data gaps can be encountered in Deliverable 2.1. Stated preference survey (SP survey) details: •Sample size planned: A sample size of 𝑛 = 500, is considered for each use cases. •Representativity: The samples must be representative of the use case’s natural population. •Response (variable) types to be collected from SP survey: •Travel behaviour and attitude •Stated preference responses containing the choices of alternatives under different scenarios (changing independent variables) •Socio-demographic metrics •Socio-economic metrics Other data (including revealed preferences) details: •Sample size: As per need and availability (e.g., census data, trip generation rates, trip distribution etc.). •Representativity: The data must be representative of the use case’s natural population matching the SP data. Application in the PHOEBE project Lack of mobility and preference data, especially in cities, is a common challenge. Consequently, surveys emerge as an indispensable and practical means of gathering data to construct these models. Anticipating the specific data needs is crucial for cities to strategize data collection efforts that not only address immediate requirements but also align with long-term planning and forecasting objectives, which are fundamental aspects of the PHOEBE framework. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Behavioural models Phoebe Applicability Individual behaviour probabilities or proportions (of traffic) at segments Adjust the behaviour distribution parameter in microsimulation Discrete Choice Model P(Behaviour) = f(βX) X = geometry, traffic, time of day, enforcement, SR, sample attributes Star Ratings (SR) of segments Instances of a specific road user behaviour iRAP Road data Questionnaire and survey data Use cases Geometry, traffic composition Telematics FLOOW,O7 Run simulation Behavioural models developed based on the data and processes explained in slides previous slides will be integrated into the Aimsun microsimulation models. The following flow chart elaborates the steps involved in the integration process and shows how behavioural modelling component will interact with microsimulation component within the PHOEBE framework. Further details are provided under Section 3 (Integration Protocols). This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Behavioural Models Zero-inflated Combined Beta Models If combined data from surveys and telematics (stated choices and revealed data) are to be used, zero-inflated combined beta models will be the suitable models in this case. The zero-inflation is suggested as we also have zero responses (e.g. no speeding) at some segments in the telematics data source. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Behavioural Models Logit models for survey (+SP) If only survey data are to be used then logistic regression are suggested, as follows: This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Behavioural models: Design of Behavioural Survey The behavioural survey will consist of two parts, including: •Questions on demographics, past behaviours and personality traits, and •Questions based on stated preference. It will focus on the study behaviours. The total number of questions determined based on number of applicable scenarios considering the key influencing factors. Application in the PHOEBE project The following behaviours will be studied within each use-case. Study behaviours within the use-cases are based on the prioritised behaviours and data availability for the use-cases. 40 ATH Speeding Non-compliant behaviours of pedestrians VLC Speeding Non-compliant behaviours of micro-mobility users WM Note: Distracted driving will be studied as a R&D part of the project. Speeding This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Examples of questions related to past behaviours and personality traits. 41 Past Behaviours Behavioural models Design of Behavioural Survey Social Value Orientation This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Example of stated preference questions. Scenario: Please look at the figure, consider yourself driving your car for your daily trips on a two-way highway and answer the questions in the following scenarios: 42 Behavioural models Design of Behavioural Survey This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) The following table presents the survey requirements in terms of sample size and targeted audience, per use case and behaviour type. 43 Survey Use case Behaviour Sample size Time Mode #1 Athens Speeding 500 ~ 40 mins Car drivers #2 WM Speeding 500 ~ 40 mins Car drivers #3 Valencia Speeding 500 ~ 40 mins Car drivers #4 Athens Pedestrian compliance 500 ~ 40 mins Peds #5 Valencia Micro-mobility compliance 500 ~ 40 mins Micro-mobility users Behavioural models Behavioural Survey requirements This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) iRAP’s Star Rating safety assessment models (including both the Star Rating and the Fatal and Serious Injury (FSI) Estimation models*, have been updated to align with PHOEBE principles: •Improving predictive safety capability for urban road networks – VRU safety driven •Model optimisation for higher quality, AI and ML-derived data The changes improve the functioning of the models for pedestrian and bicyclist safety in urban areas.This includes, among other things, improved ability to model safety benefits/disbenefits of smaller changes in traffic speeds (especially at low speeds), ability to better model paths shared by bicyclists and pedestrians, and the safety of mid-block crossings for bicyclists. The changes also bring the models in closer alignment with Safe System principles, particularly around safe speeds for where pedestrians and bicyclists are mixing with vehicles (shared streets, intersections and crossings). The following slides describe the key changes to the models and how the changes affect how data is recorded for the models.The full data specifications are in the updated iRAP Coding Manual and risk factors in the iRAP Risk Factor Factsheet. 44 Road Safety Assessment Models: PHOEBE Enhancements 0 0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100105 11 0 115120 12 5 130135 14 0 145150 Risk factor km per hour New speed risk curves for pedestrians and bicyclists (compared to previous) Bicyclist (previous) Pedestrian (previous) Bicyclist (new) Pedestrian (new) *iRAP’s Star Rating and the Fatal and Serious Injury (FSI) Estimation models are interrelated predictive road safety assessment models.Star Ratings show the individual risk of the road, while FSI Estimation models uses the Star Rating, the numbers of road users (by mode) and records of past FSI crashes to show the collective risk, that is, the overall number of expected FSI crashes over a given period (e.g. 20 years). See Deliverable 1.2 for more information. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Pedestrian Star Ratings Main changes: •1 km/h incremental speed •Review on the speed risk factors •Change in the crossing and along risk factors •More crossing and intersection categories •Sidewalk and crossing quality •Star Ratings based on operating speeds •Decimal Star Ratings Bicyclist Star Ratings Main changes •1 km/h incremental speed •Review on the speed risk factors •Introduction of a new crash type – Crossing •Review of contributing factors and risk scores •Star Ratings based on operating speeds 45 Road Safety Assessment Models: PHOEBE Enhancements The main improvements in the pedestrian and bicyclist road safety assessment models are described below: Alignment with global road safety goals This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Along risk + No. of lanes (more lanes = higher risk) + Paved shoulders (present = lower risk for sidewalks) + Property entrances (present = higher risk) + ‘Informal’ sidewalk categories have been changed to poor and medium quality categories Crossing risk + No. of lanes, lane width, paved shoulders and intersections with turn lanes are used to better estimate crossing distance + Specific speed related risk factors for pedestrian crossings + More crossing and intersection categories 46 Road Safety Assessment Models: iRAP Pedestrian Model Enhancements The iRAP pedestrian risk model calculate the risk when pedestrians cross the road (inspected or side road) and when walking along the road (driver side and passenger side). The main changes in each of the model components is described below. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Models Extracting mean speed and 85th percentile from simulation 53 User interface to generate the databases In red, the settings to export the average speed and standard deviations for every hour. It is very important to set the statistics gathering time to the appropriate time duration, in this case 1 hour. In blue, the tick boxes to generate the vehicle trajectories per section. If Aimsun Next statistics are used, there is the need to extract the average speed, standard deviation, count and time. This is done doing a query like the following to the SQLite database. SELECT oid, ent, count, speed, speed_D FROM MESECT The query returns the data from the database, which then it needs to be further processed by each section (oid) and hour (ent) to get the average speed (speed) and the 85th percentile speed (speed+1.036*speed_D). The count is to ensure that the sample is large enough to be representative. As per iRAP standards a minimum of 100 vehicles are needed take the sampling as valid. To inspect the SQLite file produced by Aimsun is recommended to use a SQLite software to inspect the fields available. In particular DB Browser for SQLite is a lightweight application that does very well the job https://sqlitebrowser.org/. To process the data in python the packages sqlite3 and pandas are used. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Models Dynamic risk assessments Dynamic flows Two approaches are going to be tested. The first is the expansion factors to transform hourly flow into AADT. Pros: •Customized for each road/network. Cons: •It demands traffic historical data for identification of the hourly flows expansion factors. •Reduced variability of the data leading to a smaller influence on the traffic flows effects on the risk results. 54 Impact test will be conducted upon receiving information about the use cases simulation scenarios Example from S108 Queensway (West Midlands) -4 lanes Road AADT based on 365 days counting data: 64,671 _ AADT per lane = 16,168 (Only car, motorbike and bus countings were available for all days) Typical day: 14/02/23 Total flow of the day: 69,594 Morning off peak (OPH) = 2,608 (3.7% of the day flow) Afternoon peak hour (PH) = 4,859 vehicles (7,0% of the day flow) Transformation to peak road AADT/AADT per lane: OPH to Road AADT using 3,7% expansion factor : 70,486/17,621 PH to Road AADT using 7,0% expansion factor : 69,414/17,353 Star Ratings risk factor category = 19 (17,000 to 17,999) for both cases ( Risk factors values area available at iRAP methodology factsheets) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Models Dynamic risk assessments Dynamic flows Two approaches are going to be tested. The second is transformation of daily risk bands into hourly risk bands. Pros: •It doesn't need historical data. •Ensure higher variability of risk. Cons: •Generic for urban context. 55 Example from S108 Queensway (West Midlands) -4 lanes Band transformation considering overall 10% expansion factor: Morning off peak (OPH) = 2,608 Afternoon peak hour (PH) = 4,859 vehicles OPH per lane = 652 vehicles PH per lane = 1214 vehicles OPH risk category = 8 (600 to 699) PH risk category = 14 (1200 to 1299) Impact test will be conducted upon receiving information about the use cases simulation scenarios This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Socioeconomic analysis The socio-economic analysis (SEA) will be carried out to appraise the changes introduced in the system. The steps are: •Fixing Key Performance Indicators (KPI) based on which the SEA will be carried out. •Scaling up the all the modelling and analyses outputs at the network level. •Carrying out a Cost-Benefit Analysis (CBA). •Establishing a feedback loop with the former modelling and analyses procedure based on the output of the CBA. The last step of PHOBE Framework (see Fig 2-1, D1.2) of the framework is the SEA). Through SEA, the efficacy of the changes will be determined. The SEA, consists of a Cost-Benefit Analysis (CBA) at its core. The SEA will establish a feedback loop that will control the changes administered in the system in Step 2 (Fig 2-1, D1.2). 56 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Socioeconomic analyses Scaling up procedure The simulations and modelling procedures of PHOEBE will produce outputs in the forms of KPIs corresponding to a few components of the whole network. These outputs will be scaled up to the whole network level of the use-cases to estimate the overall impacts due to the changes. The scaling up procedure envisioned for this purpose is explained as follows. 57 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 58 Socioeconomic analyses Cost Benefit Analyses (CBA) The network loaded with changes and impacts will be used to estimate the costs and benefits. Vital parameters, e.g., Value of Time*, Value of Preventing Fatality*, Willingness to pay* etc. will be estimated from the coefficients of the deterministic utilities of the discrete choice models (mode choice and induced demand). 𝑽𝒏𝒊 =𝜶𝒊𝟎 +𝜶𝒊𝟏.𝒙𝒏𝒊𝟏 +𝜶𝒊𝟐.𝒙𝒏𝒊𝟐 +⋯ Value of Time (VoT): The value of time (VoT) is the opportunity cost of the time that a traveler spends on their journey. In essence, this makes it the amount that a traveler would be willing to pay in order to save time, or the amount they would accept as compensation for lost time. Value of Preventing Fatality: There are several metrics for calculating or estimating the benefits of road safety measures. One such metric is the value of statistical life (VOSL). VOSL is also called the value of life, the value of preventing fatality (VPF) or the implied cost of averting a fatality (ICAF). Willingness to pay: Willingness to pay, sometimes abbreviated as WTP, is the maximum price a customer is willing to pay for a product or service. It’s typically represented by a dollar figure or, in some cases, a price range. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 59 Socioeconomic analyses Example and next steps The SEA is the last component of PHOEBE and depend on the results of all the models in the before and after scenarios. Therefore, the content present in D3.1 serve as aguideline on how SEA analysis will be performed and tested in the PHOEBE framework. Example: The estimation of the toy model (mode choice) renders the following coefficients: •𝛼,-. =1.536 •𝛼/=−2.638ℎ𝑟0) •𝛼,=−0.215$0) The Value of Time (VoT) can be calculated as: •𝑉𝑜𝑇= 1! 1"⁄ $ℎ𝑟 =12.27$/ℎ𝑟 Based on the results of the CBA, a feedback loop is to be established through which the changes can be fine-tuned (see Deliverable 1.2). Application in the PHOEBE project As the culminating component of the framework of PHOEBE, socioeconomic analysis will appraise or measure the effectiveness of the changes introduced in the system to attain the stipulated objectives of PHOEBE—the sequence of the SEA as a takeaway is presented below. •Scaling-up operation:Since the PHOEBE framework is designed to achieve the stipulated objectives through a microscopic approach, a complete simulation of the use cases is out of the question. In order to solve this, a scaling-up operation will be carried out to extrapolate the microscopic simulation results for the entire domain of the use cases. •KPIs:A comprehensive list of the KPIs will be synthesized (refer to Deliverable 1.2 and 4.1)to capture the outputs of the modelling components of the PHOEBE framework.The scaling-up procedure will extrapolate these KPIs to the entire extent of the use cases. •Socioeconomic Analysis:At the core of the SEA exercise lies the Cost-Benefit Analysis (CBA).Through CBA, the changes' efficacy/effectiveness/appraisal will be calculated. •Feedback loop:Based on the outputs of the SEA (CBA), a feedback loop will be set up that will be detrimental to adjusting the potency/efficacy of the incorporated change to ensure road safety. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 3) Integration protocols Integration of PHOEBE Components into Traffic Simulation This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) About Chapter 3 61 Road Safety Assessment models Behaviour models Demand models The traffic simulation model are the receptors for all components, and it is presented transversally throughout the chapter. This chapter delineates the progress made up to this point in the project, focusing on the integration of PHOEBE components into traffic simulation. The integration protocols will be completely developed by the end of WP5. The chapter content is organised by model component (demand models, behaviour models, safety models respectively) and specifies/describes the following aspects for each if relevant: •The communication protocols used for sending data t othe model and receiving predictions. •Theprocedures for deploying the integrated model to production. •The format of input data and the expected format of output predictions. •Examples of requests and responses. •Pre-description of API. How to navigate this chapter Click on the files on the right to access a specific subsection of this document, or progress through the slides to read the whole chapter. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This is a simplified version of PHOEBE workflow to show interaction staged between demand models and traffic simulation. Detailed version is available at Deliverable 1.2. Run simulation Export data though standard Aimsun procedures and code if necessary Post simulation analysis Simulation results: •Speeds •Trajectories •… Initial OD matrix Mode choice is fine and converges New OD matrix No New matrix introduced though scripting Model is safe? No Yes Yes DONE 62 Demand models Inputs-outputs of mode choice in traffic simulation This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Behavioural models Probability function representing pedestrian non-compliant behaviour – crossing away from designated crosswalk 71 Input features Conditions for obtaining the highest probability x1 = 18 # Age x2 = 1 # Gender x3 = 1 # Group x4 = 1 # No lanes x5 = 1 # Distraction x6 = 7 #Gap size (aka TTA/TTC) x7 = 1 #Intention to cross Conditions for obtaining the lowest probability x1 = 80 x2 = 0 x3 = 0 x4 = 3 x5 = 0 x6 = 2 x7 = 1 Optimal values after scaling: a = 0.8383438011104125 b1 = 0.5755838131330817 b2 = 1.0910175557620763 b3 = 1.0910175557620763 b4 = 0.3329965544962949 b5 = 1.0910175583619688 b6 =7.131775526798777 b7 = 0.8383438273307598 Probability for highest probability condition: 1.0 Probability for lowest probability condition: 2.284459614468522e13 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching Challenge: Develop a robust data matching procedure to streamline the integration of data from various sources and harmonize networks with disparate data structures. Main Objective: Ensure that risks are accurately represented within the simulation network, reflecting the real conditions of roads or scenarios under consideration.This involves considering geographical location, operational attributes, and other pertinent factors. Secondary objectives: I. Automate the process as much as possible to to minimize processing time and guarantee standardized results across diverse datasets. II. Maintain direct links to the original datasets, enabling thorough verification and validation of the matching procedures to ensure accuracy and reliability. III. Investigate alternative data configuration methods to improve matching accuracy. Risk evaluation iRAP length-based assessment where each row in the database represents 100m segment Traffic simualtion AIMSUN network structured by sections connected by nodes. Nodes are point or an area in the network where sections are linked and where vehicles move onto their next section 100m Telematics data Telematics data is spatially represented by connected links. 78 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching - Procedure for links 1 Generate links to represent 100/10-meter segments within the dataset. This step is essential because the downloadable database from iRAP is configured with data points representing the beginning of each 100-meter segment. 2 Generate links centroids These centroids are necessary for creating the influence areas of the data, as outlined in the subsequent step. 3 Generate data influence zones by creating buffer zones around the centroids. These buffer zones will encapsulate all the information from the links/centroids and will be utilized for spatial matching with the traffic simulation sections. 79 The same procedure was utilized to connect telematics data and can also be applied to other link-based data sources from platforms like OpenStreetMap. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching - Procedure for links 4 Perform spatial matching The spatial matching will attribute buffer data to all the links that intersect the influence area. 5 Match automated validation During this step, we assess whether the road name and the number of lanes match precisely between the buffer and the sections. Any features that do not have a perfect match, such as a discrepancy in the number of lanes or road name, will require further review and revision. 6 Match manual revision Perform manual revision of the matching process by color-coding to identify two categories: (i) sections missing information, and (ii) sections incorrectly matched. 80 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessments Network matching High complexity intersections Mismatching coordinates Dual carriageways overlapping Imprecise design of the network 83 Examples of sections that need manual revision: This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching 84 As previously mentioned iRAP methodology indicates a length based assessment where road-related attributes are coded every 100m. This level of granularity is suitable for capturing risk variability in most road networks. However, in dense urban centers where blocks are shorter than 100 meters, this approach can pose challenges. The primary recommendation is to code the most hazardous situations within a100meter stretch, potentially leading to crucial data omissions. This issue is particularly evident in the context of intersections. Consider the example in the Athens area for illustration: Two intersections are part of the 100min section B: a 3-leg intersection and a 4-leg intersection. By coding solely within the 100-meter intervals, only the 4-leg intersection would be identified in the coding database. Hence, further examinations were undertaken utilizing existing data for the WM areas, aiming to compare the network matching accuracy between 100-meter segments and 10-meter segments. The results of these tests are outlined in the table below. 47m 62m 60m A:100m B:100m Accuracy parameters 100m database 10m database % of data correctly matched using the automatic approach (See here) 59.7% 80.4% % of data matching manually adjusted 40.3% 19.6% 10m coding structure improve data matching ATH 100m vs 10m segmentation This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching - Procedure for nodes 1 Search for intersections in the risk database At this stage, it is only pertinent to work with the data points in the risk database where intersections were identified during data collection and intersection risks were calculated. 2 Access to the node shape The node shaped need to be extracted from AIMSUN (see more here). Attribute name of the road and number of lane from the connecting link information can help to improve matching accuracy. 3 Associate nearest intersection risk data to the nodes These nodes shapes are the global representations of all turns in an intersections. 85 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 86 Accuracy parameters 100m database % of data correctly matched using the automatic approach (See here)49.1% % of data matching manually adjusted 50.9% Road Safety Assessment Network matching - Procedure for nodes As the results indicate, matching nodes poses more challenges than matching links in the network.The increased percentage of manual adjustments can be attributed to the significant presence of roundabouts, which are modeled with nodes at each entrance and exit. In iRAP methodology, roundabouts are coded as a single entity, and the risk is calculated for the entire intersection. 3 Therefore, to align with iRAP standards, the intersection risk must be manually assigned to each node that comprises the roundabout.Below is an example illustrating this process. The results of the node accuracy tests are outlined in the table below. WM This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching –Shape extractions in AIMSUN 87 Nodes as seen in Aimsun Next Nodes shapes are extracted by using the function getMinimalPolygons() of the object instances GKNodes. This provides a vector of Aimsun points (GKPoints) with the geographical extension of the turns that composes each node.Then, further processing the points into python objects and computing the external boundary enables to export the node shapes from Aimsun into a shapefile with the node ID. GIS file outputs viewed in QGIS This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Network matching – Incorporating risk data in AIMSUN Application in the PHOEBE project Once the iRAP database and Aimsun sections are matched, it is possible to generate aTable with the iRAP safety assessment values that correspond to each of the Aimsun network.They need to be added into Aimsun section as custom attributes.This is done adding extra fields to the existing Aimsun objects using the extensible metadata system and can be done through code or manually.For PHOEBE integration is done through code, using the scripting library. 1. The file containing iRAP road risk assessments (csv file) is read using pandas module in python (https://pandas.pydata.org/) 2. The column names are used as Aimsun variables with spaces removed 3. One custom attribute in Aimsun section objects is created per each variable from iRAP that needs to be added into the simulation model.The attribute is generated with adequate variable type (integer, float or string) and the adequate level of precision, which for decimal values is up to 10 decimal places. 4. For each section and variable, the values are written into the simulation model. 88 Section object type in Aimsun Next iRAP risk assessment file This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Demand model Calibration and validation The development of the trip distribution and mode choice models include several processes: •Model Estimation: Statistical estimation of the models from the observed data. •Model Calibration: The process of adjusting the model specification and/or model parameters to achieve validation is referred to as calibration. •Model Validation: Validation refers to the comparison of model results to independent observed data not used in parameter estimation. Methods or measures used for calibration •Comparison of forecasted and observed OD matrices •Sensitivity analysis •Trip length/frequency/time distribution •Intrazonal trip percentages •Aggregate District-to-District flows •Screenline counts •Realism tests 96 Adapted from: Pestel, E., Friedrich, M., Heidl, U., Pillat, J., Schiller, C., & Schimpf, M. (2016). Qualitätssicherung von Verkehrsnachfragemodellen. Straßenverkehrstechnik, 60(10). This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Stated Preference Surveys Demand Models Data requirements for validation In order to validate the demand models, the data below is required. •Traffic flow data •Traffic count data •Actual mode choice data •Actual VKT counts Data sources for the required data: 97 Telematics Road user counts Mapping initiatives This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) For demand model estimation, python package Biogeme/R package Apollo will be used. 100 Demand model Support Software This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Behavioural models Validation procedures 101 The development of the behavioural models also include several processes: •Model Estimation: Statistical estimation of the models from the observed data. •Model Calibration: Verification of model fitting parameters •Data validation: Use of real data observations. •Model Validation: Part of the model development process (e.g, use of kfold cross validation) To validate the behavioural model, it is necessary to obtain observed behavioural data. Data sources for the required data: Stated Preference Surveys Telematics Video data This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Model Pedestrian model enhancements - Real life testing 103 The enhancements to the Star Rating Models for pedestrians and bicyclists have been meticulously developed and are prepared for testing within the PHOEBE framework across the three use cases. It's important to note that the PHOEBE components, more specifically the risk assessments and traffic simulation, have garnered global usage. Consequently, the pedestrian model enhancements have already undergone real-world testing in diverse locations worldwide. The images provided below showcase some of the outcomes derived from the model, demonstrating its heightened sensitivity to inadequate infrastructure quality while also acknowledging and rewarding well-designed pedestrian environments. These updates to the model make it more rigorous, requiring improved infrastructure to attain a 5-star rating. India Zambia India This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Traffic simulation model Calibration/validation Models are calibrated comparing the simulation results with real data sets (RDS).This is data from road sensors that provide at minimum flow, ideally also occupancy and speeds averages over regular periods, typically in the few minutes region. The metric used is the GEH statistic, which is aformula to compare different traffic data sets. In this case the measured data (𝑜)and simulated data (𝑚): 𝐺𝐸𝐻 = 2 𝑚 − 𝑜 # 𝑚 + 𝑜 Awell calibrated model will result in a small value of GEH while apoorly calibrated will have a high GEH. To improve GEH in a microscopic simulation model, expertise is needed to adjust the parameters that critically affect the high GEH among the tens of parameters available. This is performed on the baseline scenario (do nothing) to ensure that the model will be reasonably accurate when the prediction is done. 104 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 5) Users documentation Information available at project month 18 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) About Chapter 5 106 Road Safety Assessment Model Traffic Simulation Model This chapter initiates the compilation of materials intended for inclusion in the PHOEBE user support documentation. The content gathered here will serve as afoundational resource to assist cities in effectively implementing the PHOEBE framework. Assembling comprehensive and userfriendly documentation is essential to providing users with the guidance and assistance needed to optimize their experience with PHOEBE. The chapter content is organised by model component (traffic simulation and road safety assessments) and and specifies/describes the following aspects for each if relevant: •Guidelines on how toc reate new views in AIMSUN that allow risk visualization. •Update iRAP guidelines based on the enhancements described in Chapter 2 User documentation for the behaviour models and socioeconomic model will be availabel after the model developements are completed (after Milestone 3 is achieved). Demand Modelling Behaviour Models How to navigate this chapter Click on the files on the right to access a specific subsection of this document, or progress through the slides to read the whole chapter. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 107 Demand modelling User documentation In this section a sequence of the steps are provided that may be followed to develop relevant demand models (mode choice models and induced demand models). Please refer to the figure in the next slides for a visual representation of the steps. Delineating the data requirements For estimating demand models two sets of data are required representative to the use case population: a. Data for model estimation: Stated preference data (attitudes from surveys), Revealed data (e.g., flow (AADT), cost (fuel/ticket prices), OD volumes, speed, networks, land use) b. Data for model calibration and validation: Primarily revealed data (e.g., flow (AADT), counts in terms of VKT/VMT, PassengerKM, OD volume) Model estimation, calibration and validation a. The datasets can be used to estimate the demand models (discrete choice models). b. Any software package (e.g., Pandas Biogeme,Apollo) meant for discrete choice model estimation can be used depending upon the technical proficiency. These packages are open source. c. The estimation process must be followed by suitable statistical tests to ascertain significant exogenous variables that affect the outputs of these models and different model configurations.. d. Finally, these models must be calibrated and validated with the data mentioned above in order to ensure precise recreation and forecasting. Model deployment Current state recreation:The model estimation, calibration and validation is followed by the network-wide model deployment. For this step, the whole network is needed to estimate the values of the exogenous variable for the entire network.This step will recreate the current situation, i.e., provide the current modal share/split, for induced demand model the output will not change as in base case there will be no induced demand. Predictions:These models can be used to forecast the future by changing the exogenous variables, e.g., travel times, cost. This will generate anew modal split which can be used to generate amodal shift (wrt.the current state recreation). Model integration These models can be integrated with other components, e.g., microsimulation and road safety assessment models with the help of the exogenous variables. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 108 Demand modelling User documentation: Steps in model building Steps in model building: 1. Zero configuration model:Simple models.To test the data work, Base for later configuration. 2. Functional form:Emphasis on exogenous variable types, e.g., categorical, linear, non-linear. 3. Investigation of error structure assumptions:departure from IIA assumptions, scope for heteroskedasticity across observations etc. 4. Investigation of the data (estimation):Sample coming from different sources, e.g., SP, RP, location, segmentations 5. Test for differences across segments:E.g., separate models for different segments, statistical tests for model fit differences against generic model Marginal Rates of Substitution (MRS) outputs and elasticities. 6. Ascertaining deterministic heterogeneity:link heterogeneity to observed characteristics of individual and/or scenario, gradually build up model complexity, determining confounding factors. 7. Ascertaining random heterogeneity:Determining model types, e.g., mixed logit or latent class. 8. Inclusion of Attitudes and other soft factors in the models This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessment Model Upload file specifications 115 This specification sets out the requirements for road coding data files to be uploaded into ViDA. The new version include the model enhancements describe in Chapter 2. Launch of version 3.1 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 6) Use cases dedicated models Preliminary results This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) About Chapter 6 117 Road Safety Assessment Model Traffic Simulation Model This chapter outlines the model developments and results specific to each of the use cases that are ready for demonstration. The comprehensive development of models relies on the data accessible upon the completion of Milestone 3, which marks the availability of working data for WP3, WP4, and WP5. The chapter content is organised by model component (traffic simulation and road safety assessments) and specifies/describes the following aspects for each if relevant: •Status of simulation model development •Description of strategies to generate star rating static results •Description of procedures and data used in the static results •Presentation of the preliminary results How to navigate this chapter Click on the files on the right to access a specific subsection of this document, or progress through the slides to read the whole chapter. Model results from the demand model, behaviour models and socioeconomic model will be available after Milestone 3 is achieved. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Road Safety Assessments Static Results Use cases preparation for integration Concept:The static results provide an overview of the overall road safety risk for individual users on a typical day. These results represent the baseline conditions for all the scenarios tested within the specified use cases. Differences from the Dynamic results:The static results are derived from global representations of traffic flows and speeds (such as road AADT and freeflow operating speed). Following integration, detailed data on speeds and flows generated by simulations will be fed into the risk models to dynamically calculate road safety risks across various time periods.This dynamic process enables continuous updating whenever a new scenario is tested. Different approaches are pursued in each of the use cases, contingent upon (i) the presence of existing assessments, and (ii) the configuration of the pilots and the area under analysis.The threes approach will allow projects to establish procedures for future applications of the framework. ATH VLC WM New assessments generating data every 100 meters, aligning the start and end points of the assessments with the traffic simulation network. Nodes present individual coding of the intersections. Consequently, there is no need for network matching. New assessments entail data generated every 10 meters. Network matching is necessary to merge risk assessment data with traffic simulation sections and nodes. A combination of existing and new assessments employed, with existing assessments conducted every 100 meters and new assessments carried out every 10 meters. Network matching is necessary to merge risk assessment data with traffic simulation sections and nodes. 123 This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 7) Final remarks “Confidence in the past give us the courage to look forward to the future” (Unknown author) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) What we learned Conclusions and learnings from PHOEBE developments presented in D3.1 148 Deliverable D3.1 serves as an intermediate document within WP3, showcasing the progress made in the current project stage. Its objective is to outline the work completed thus far and set the stage for the upcoming demonstration of the PHOEBE framework within the designated use cases. The project team has opted to present the deliverable in slide format, leveraging visual aids such as images, charts, and videos to effectively showcase project achievements and highlight developments. As is customary in project development, the team has gathered valuable insights that will guide our approach as we transition into the next phase. Below, we outline the key lessons learned and takeaways: The interconnections facilitated by PHOEBE significantly contribute to the complexity of the project. The theoretical framework described in D1.2 encompasses eight connections among PHOEBE components. Initially, within a first alignment layer, each pair of partners responsible for developing these connections must reach agreements regarding model capabilities to absorb and process inputs, expected outputs, data formats, and evidence supporting key decisions. Subsequently, in a second stage, the use case leaders play a crucial role in assessing the suitability of these agreements for the respective use cases and determining the specificities that need to be considered to align with the experiment designs outlined in D4.1. This process ensures that the models under development will provide PHOEBE with the necessary capabilities to accurately forecast road safety impacts. Yet, this complexity is not in vain; it is necessary due to the multidimensionality of road safety and the accurate diagnosis of its causes and impacts. VRUs remain pivotal in promoting safety within urban environments It is firmly established that vulnerable road users (VRUs) constitute the cornerstone of safety in urban environments. However, the PHOEBE project team has encountered numerous barriers in properly evaluating VRU safety. These obstacles range from the absence of dedicated data and the need to enganece models for these users to a lack of studies in specific fields.There is still much to be developed in the realm of VRU safety. The collective experience of the use case leaders, coupled with the incident data collected by the project and the preliminary results of road safety assessments, affirm PHOEBE's responsibility in addressing the needs of Vulnerable Road Users (VRUs). Data is at the core of everything. Data plays a fundamental role for PHOEBE.It is utilized in two key stages: model development and calibration/validation. It may originate from sources both internal and external to the project. Accessing, complying with requirements and processing data can indeed pose significant barriers to the reapplication of PHOEBE. Project partners are making a concerted effort to map data sources and their applicability. Guidelines on how to handle data withing the framework will support user documentation to be presented in D3.2. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) Next steps Setting the pace for PHOEBE WP3 and deliverable D3.2 149 Finalize model development Run integration trial Refine models and integration Review results with use case leaders Finalize user documentation Present outcomes in Deliverable D3.2 The upcoming phase of WP3focuses primarily on integration and prepare the documentations to PHOEBE Framework be ablueprint for cities to assess safety. Once Milestone 3 and 4is achieved and all data is accessible, model development can be finalized. With the completion of model development, the project will be poised to commence integration testing. The model and integration protocols will undergo refinement following the trial, with the possibility of requiring new interactions to be established. Use case leaders will actively participate in this process, aligning their activities with those taking place in WP4(T4.3: Analysis and assessment, T4.4: Collect local partners feedback and T4.5: Discuss results/outcomes). The model results applied to the use cases will be reviewed and shared with local stakeholders to collect feedback. Meanwhile, partners in WWP3will finalize user documentation for each of the models. These user documentations will form the central focus of Deliverable 3.2. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) 150 Next steps Setting the pace for PHOEBE WP3 and deliverable D3.2 PHOEBE is on the right track to help cities to create safe urban environments for all. This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow) This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101076963 UK participants in Horizon Europe Project PHOEBE are supported by UKRI grant numbers 10038897 (The International Road Assessment Programme –iRAP) and 10056912 (The Floow)