Recognizing the diversity of urban characteristics, street design, possible routes and user types, the research focuses on assessing the level of walkability in the city of Milan (Italy) according to the unique experiences and needs of different users. The aim is to provide valuable information for urban planners and policy makers, and to promote the analysis of more inclusive and adaptive urban environments that improve the quality of life for multiple urban users. This is based on: i) the collection of comprehensive indicators and data; ii) the construction of geospatial datasets; and iii) the development of a pedestrian mobility routing system. The three goals are supported by developing the UX Mobility Survey to gather data on users’ walking preferences, which were integrated as weights into a multi-user route planner using open-source and proprietary GIS data. The routing system results highlight how varying preferences influence UX paths, which diverge from the shortest route by prioritizing walkability factors. By applying different travel time thresholds, an optimization process is then applied to balance user experience with route efficiency, underscoring the complexity of pedestrian navigation.

Introduction

User Experience (UX) refers to the quality of interaction between a user and a system that enables the user to achieve a specific goal (Dickson-Deane & Chen, 2018). The concept of UX has received considerable attention in the field of Human-Computer Interaction (HCI) and interaction design over the past decades (Hassenzahl & Tractinsky, 2006). UX in urban design extends the principles of Human-Computer Interaction (HCI) to the built environment, emphasizing a user-centered approach to mapping and analyzing urban mobility. The interest in urban experience through participatory activities introduces a ‘user layer’ that captures the needs, expectations, feelings, and perceptions of urban users, particularly those in situations of vulnerability. Unlike traditional urban analysis, this approach collaboratively builds qualitative data layers that enrich the quantitative data traditionally used to measure urban performance (Al Maghraoui, 2019).

The challenge for the UX Mobility project is to develop a strong base of quantitative metrics that are available and open, on which subjective parameters can be built iteratively and interactively by user segments. Each road has unique quantitative and qualitative characteristics that can enhance or hinder the experience of different users (Gehl, 2010). For instance, parents with pushchairs prioritize wide, unobstructed pavements and safe crossings, while people with disabilities seek smooth surfaces, gentle ramps, and continuous footpaths.

Pedestrian navigation operates at three behavioral levels (Basu et al., 2022): strategic (i.e., pre-journey planning), tactical (i.e., route selection), and operational (i.e., real-time decisions such as crossing roads and avoiding obstacles). Route choice reflects a dynamic interplay of external influences and individual preferences, shaped by trip attributes (e.g., distance, traffic volume, and travel time), the built and natural environment (e.g., sidewalk characteristics, pedestrian amenities, safety, and aesthetics), and socio-demographic factors (e.g., age, gender, and travel companions). These factors influence both the objective selection of a route and the subjective experience of it.

This project focuses on: i) the collection of comprehensive indicators and data; ii) the construction of geospatial datasets; and iii) the development of a pedestrian mobility routing system. The three goals are supported by survey data on users’ walking preferences, which were integrated as weights into a multi-user route planner using open-source and proprietary GIS data. Within the route planner, the ability to explore different weights and record the preferences of various user segments in route calculation allows for the identification of the value of each parameter and potential navigation solutions tailored to the specific needs and priorities of each user (Bast et al., 2016; Novack et al., 2018). The final aim is thus to provide valuable information for urban planners and policy makers, to demonstrate the impact of user differentiation on route planning outcomes, and to promote the analysis of more inclusive and adaptive urban environments that improve the quality of life for multiple urban users.

Enabling Data and Methodology

This section explores the datasets and steps necessary to develop a routing system that prioritizes user preferences rather than selecting the shortest path. In particular, the approach consisted of assessing three main components:

1. The collection of comprehensive indicators and data: a survey was conducted to identify key walkability indicators and assess their importance to pedestrians. Respondents were then grouped into clusters based on their priorities;
2. The construction of geospatial datasets: various data sources were used to quantify static factors relevant to walkability and integrate them into the street network;
3. The development of a pedestrian mobility routing system: indicators gathered through a survey and geospatial datasets on the street network were integrated into a routing system that considers walkability factors in the calculation of the optimal path.

The collection of comprehensive indicators and data

The first phase of the project involved the collection and selection of key indicators of walkability related to multiple users. The basic parameters for evaluating these indicators are based on guidelines developed by the Walk21 Foundation, for the Walkability App, a measuring tool developed by the globally recognized organization with the aim of promoting and sharing walking experiences. These indicators, divided into categories (level L1) and subcategories (level L2), identified as relevant within the pedestrian experience, provided a solid framework for evaluating different mobility scenarios.

To further explore these indicators, the proposed UX Mobility Survey was launched in July and closed in December 2024, to gather comprehensive data on UX factors that influence the pedestrian experience (see Figure 1). The survey collected 290 answers, with the sample consisting of 12% under the age of 25, 61% between 26 and 35 and the remaining 26% older. Of these, 57% women, 40% men and 3% other, 40% were of Italian citizenship and around 30% residents in the city of Milan, with a total of 5% bearers of a disability. The survey explored the experience of participants as city users, providing insights on their walking habits, primary transportation modes, purpose of walks, duration and mode (whether people usually walk alone or accompanied). The main section delves into the analysis of the specific indicators influencing the walking experience, presenting each Walkability Indicator and providing a scale of 1 to 5, 1 being not at all relevant and 5 being very relevant. Finally, the survey collected background information on respondents, including age, gender identity, citizenship, disability status, and occupation.

After conducting the survey, the collected data was analyzed to identify patterns in walking preferences and group respondents into clusters. The goal of this clustering analysis was to determine the optimal number of user groups based on the importance assigned to Walkability Indicators, as reflected in the survey responses. From the 33 indicators assessed in the survey, 12 key indicators were selected and converted into data for analysis.

The K-means clustering method, an unsupervised learning algorithm, was used to identify distinct user groups by assigning data points to k clusters and iteratively adjusting centroids for optimal grouping. The ideal number of clusters was determined using the elbow method, which analyzes the within-cluster sum of squares and selects the point where the decrease rate slows. The analysis identified four distinct user groups based on the importance ratings of the 12 Walkability Indicators. Then, an analysis of clustering results was performed to create weights to be used as inputs in the routing model.

To generate meaningful weight values for routing, a multi-step normalization and standardization process was applied. First, the average values of key indicators were calculated for each category. Then, data was scaled using min-max normalization to ensure comparability across different metrics. Next, Z-score standardization adjusted values based on their distribution. Finally, an exponential transformation was applied to refine weight influence and maintain positive values, ensuring smooth integration into the routing model.

The construction of geospatial datasets

The second phase of this project involved the construction of a geospatial dataset, mapping the 12 Walkability Indicators analyzed to derive clusters on the street graph of Milan (Italy) (see Table 2). This street map was constructed using the open-source data of OpenStreetMap and then updated and refined using Geographic Information Systems (GIS). The different indices were integrated into the graph according to specific methodologies based on the index typology. Percentile-based scaling is applied across all indicators, penalizing extreme values such as narrow sidewalks or high pollution levels while prioritizing more favorable conditions. Indicators are normalized to a scale between -1 and 1, with the median value set to 0, allowing for standardized comparison across different metrics. This integration ensured that the map reflects the various indicators by providing a solid basis for the development of the routing system.

The development of a pedestrian mobility routing system

In the third phase of the project, a Python-based routing system was developed to generate customized paths. The process begins with graph construction, where a street network shapefile is converted into an undirected graph. Nodes represent intersections, while edges represent street segments enriched with geospatial datasets of the 12 Walkability Indicators described in the previous section.

Next, user-defined start and end points are mapped by identifying the nearest graph nodes. To account for user preferences, cluster-based weights, described above, are imported, reflecting different priorities in route selection.

The system then computes two alternative routes: one based on shortest distance and another optimized for user experience (UX). The UX route is determined by a custom weight function that integrates user preferences with street attributes. The cost of a segment is determined by multiplying the presence of different walkability factors mapped on the street segments (i.e., geospatial datasets) by their assigned importance for each user cluster (i.e., Walkability Indicators). The final route is then calculated by summing these costs across all segments, guiding pedestrians along paths that best match their preferences and priorities.

The code also includes steps to balance distance in UX path calculations. First, an optimal path search is performed through an iterative process that adjusts an alpha parameter to find a balance between the shortest and UX-optimized paths. This generates routes that are approximately 5-10%, 10-25%, and 25-50% longer than the shortest path while preserving user-defined weights.

Last, the final path is generated, storing and visualizing multiple routes, including the shortest path, the UX-weighted path, and three balanced alternatives. Additional diagnostic features provide insights into route performance, such as the number of crossings, sidewalk width, and traffic volume, aiding mobility decisions.

Results

This section presents the results for the Milan case study, focusing on a quadrant of the city that includes the area around Piazza Duomo to the northeast and Piazza Piola. It details how respondents of the UX Mobility Survey were grouped into four clusters based on their walking preferences, which were defined using 12 Walkability Indicators. It also examines the spatial distribution of geospatial datasets quantifying the presence of these indicators in the street network. Finally, it outlines the combined results of integrating user weights and datasets into the routing system, highlighting the suggested paths for each cluster.

Figure 2 outlines the results of the k-means clustering process. The analysis identified four distinct groups based on respondents’ preferences regarding 12 Walkability Indicators. These clusters were then analyzed to determine which elements were most important to each group and derive a semantic description of their user persona (see Figure 2).

Figure 3 illustrates the spatial distribution of the 12 Walkability Indicators from Table 2, mapped onto the street network within the case study area. The combined presence or absence of these indicators in the network, along with the cluster weights, determines the final scoring of each street segment in the routing system.

Figure 3 Spatial distribution of 12 Walkability Indicators

Figure 4 presents the results of the routing system algorithm for the Milan case study, featuring a sample path with its origin at Piazza Piola and destination at Piazza Duomo. The routing system integrates user preferences with street attributes by multiplying the cluster weights with the level of presence of each Walkability Indicator on each street segment and summing the results for all 12 indicators. As shown, each cluster prioritizes different attributes, resulting in different paths. While the shortest path is more direct, the trade-off between UX variables leads to longer routes. Each route offers a unique combination of advantages and compromises, reflecting the complex factors that influence urban navigation.

Figure 4 Routing system results – clusters paths between Piazza Piola and Duomo

Figure 5 illustrates the optimized paths for Cluster 4, demonstrating how the time threshold significantly impacts the results and highlighting the complexity of balancing various walkability factors within a real road network. The findings indicate that while optimizing time thresholds and reducing path lengths, the presence of Walkability Indicators remains balanced and, in some cases, even exceeds that of the UX path. Table 3, meanwhile, compares the lengths of the optimal UX path and the balanced alternatives.

Figure 5 Time optimization thresholds result for Cluster 4

Conclusions and Future Works

The outcomes of the UX Mobility project concern both practical and theoretical advances in mobility and urban design. Using a user-centered approach, this study aims to advance the current analytical model of the city by providing advanced insights into the needs and preferences of different user groups. In particular, the project aimed to support the development of more inclusive and adaptable urban environments and provide valuable information for urban planners and policy makers in developing solutions to improve urban mobility infrastructure. By identifying and addressing the specific needs of urban users, we aim to bridge the gap between quantitative data and subjective user experiences, promoting a more holistic approach to urban mobility.

To achieve these goals, a survey was conducted to assess the importance of Walkability Indicators developed by the Walk21 Foundation. Data from 290 respondents to the UX Mobility Survey were analyzed and grouped into four clusters, to derive weights that measure the relevance of each indicator. These weights were then integrated into a routing system that combines them with geospatial data representing the indicators on the street network. The routing system generates different routes for each cluster, which deviate from the shortest path by maximizing and balancing indicator presence according to user preferences. An optimization process is then applied to refine the routes, making the tool more realistic by considering the importance of path length in route selection. It highlights the complexity of balancing various walkability factors within a routing system based on a real road network.

Future developments will focus on expanding the database of Walkability Indicators, refining the weighting function, and testing whether the generated routes align with stated user preferences. Additionally, efforts will focus on improving data collection methods and refining the survey methodology to ensure more accurate and representative results. This includes optimizing survey design, enhancing question clarity, and incorporating interactive elements to better capture user preferences. Expanding the user base will also be a priority, allowing for a more diverse range of inputs and a broader understanding of walking behaviors across different demographics.

Acknowledgments

We are grateful to the Walk21 Foundation for supporting this research and sharing information on key indicators of walkability. We thank Comune di Milano, Cittadini per l’Aria Onlus for their fruitful collaboration and for sharing data. Furthermore, we sincerely thank the respondents of the UX Mobility Survey for their valuable participation, time, and willingness to share their walking needs with Transform Transport. The data analyzed was treated in accordance with the GDPR-General Data Protection Regulation (EU, 2016/679). This research did not receive any specific grant from any funding body in the public, commercial or non-profit sectors.

References

Al Maghraoui, O. (2019). Designing for Urban Mobility-Modeling the traveler experience (Doctoral dissertation, Université Paris Saclay, COmUE). Available at: http://www.theses.fr/2019SACLC006/document

Bast, H., Delling, D., Goldberg, A., Müller-Hannemann, M., Pajor, T., Sanders, P., … & Werneck, R. F. (2016). Route planning in transportation networks. Algorithm engineering: Selected results and surveys, 19-80. https://doi.org/10.1007/978-3-319-49487-6_2

Basu, N., Haque, Md. M., King, M., Kamruzzaman, Md., & Oviedo-Trespalacios, O. (2022). A systematic review of the factors associated with pedestrian route choice. Transport Reviews, 42(5), 672–694. https://doi.org/10.1080/01441647.2021.2000064

Dickson-Deane, C. and Chen, H.L.O. (2018). Understanding User Experience. In: Encyclopedia of Information Science and Technology, Fourth Edition (pp. 7599-7608). IGI Global. https://doi.org/10.4018/978-1-5225-7598-6.ch117 

Gehl, J. (2010). Cities for People. Washington, DC: Island Press. Available at: https://archive.org/details/cities-for-people-jan-gehl/page/n3/mode/2up

Hassenzahl, M. and Tractinsky, N. (2006). User Experience – A Research Agenda. Behaviour and Information Technology, 25(2), 91-97. https://doi.org/10.1080/01449290500330331 

Novack, T., Wang, Z., Zipf, A. (2018). A System for Generating Customized Pleasant Pedestrian Routes Based on OpenStreetMap Data. Sensors. 2018; 18(11):3794. https://doi.org/10.3390/s18113794