AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations

AI-based simulation model for optimal placement of microhubs and cargo bike pick-up stations Deliverable 2.3 Version 2.0 Project title: Fostering sustainable consumer behaviour with inclusive bicycle logistics infrastructure in urban outskirts Project acronym: SuCoLo Project duration: 01/2024 – 06/2026 Project number: F-DUT-2022-0007 Work package/Task: WP2 / T2.3 Project website: https://sucolo.eu/ Authors: Benjamin Gaunitz, Silvia Torres Landaverde, Viola Süß (ULEI) This project has been funded by the Austrian Research Promotion Agency (FFG), Ministry of Enterprises and Made in Italy (MIMIT), the Federal Ministry of Research, Technology and Space in Germany (BMFTR) and the Swedish funding agency (Vinnova) under the Driving Urban Transitions Partnership, which has been co-funded by the European Union under grant agreement no. 905465. D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 2 Document versions Version Date Changes Authors V0.1 14.01.2025 Initial document V. Süß (ULEI) V0.2 05.03.2025 Documentation Simulation Model V. Süß (ULEI) V0.3 16.04.2025 Documentation Simulation Model V. Süß (ULEI) V0.4 25.04.2025 Documentation Simulation Model V. Süß (ULEI) V0.5 28.04.2025 Document revisions & finalization S. Torres Landaverde, B. Gaunitz (ULEI) V0.6 30.04.2025 Revision V. Süß (ULEI) V0.7 12.06.2025 Proofreading M. Thelen (SRFG) V1.0 04.07.2025 Final draft V. Süß (ULEI), M. Thelen (SRFG) V2.0 10.10.2025 Updates Implemented, 2nd App Version V. Süß (ULEI) List of abbreviations AI Artificial Intelligence D Deliverable ETL Extract, Transform, Load FLP Facility Location Problem GIS Geographic Information System GUI Graphical User Interface LIS Leipzig Information System ML Machine Learning OSM Open Street Map POI Point of Interest T Task D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 3 Table of contents Administrative information ..................................................................................................... 5 Purpose of the document ...................................................................................................... 6 Executive Summary .............................................................................................................. 6 1. Motivation and background ............................................................................................ 7 2. Problem description and model formulation .................................................................... 7 2.1. Model assumptions, data and parameters for Leipzig ................................................. 8 2.2. Input data for Leipzig .................................................................................................13 2.3. Input data for Merano .................................................................................................13 3. Simulation Model ...........................................................................................................14 3.1. Data preparation ........................................................................................................14 3.2. Model Framework ......................................................................................................15 4. User interface of the “Location Finder App” ...................................................................16 4.1. Model Evaluation and further implementation .......................................................21 4.2. Case Study in Germany, Leipzig ...........................................................................21 5. Discussion and further work ..........................................................................................22 References ...........................................................................................................................23 D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 4 List of Tables Table 1 Example of Digital Personas for Model Use .............................................................. 9 Table 2 User stories for bike courier company users of the app ...........................................11 Table 3 List of potential Features .........................................................................................12 Table 4 Input Data from LIS (5) ............................................................................................13 Table 5 GUI functions ...........................................................................................................18 List of Figures Figure 1 Three personas based on target groups .................................................................. 8 Figure 2 Simulation Model Framework .................................................................................15 Figure 3 Simulation model framework ..................................................................................16 Figure 4 Screenshot of the app’s start screen ......................................................................16 Figure 5 Demonstration of the simulation model results based on set preferences ...............17 Figure 6 Screenshot of the app centred in Leipzig ................................................................21 D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 5 Administrative information Basic information on the SuCoLo project and this deliverable: Project title SuCoLo: Fostering sustainable consumer behaviour with inclusive bicycle logistics infrastructure in urban outskirts Project coordinator Salzburg Research Forschungsgesellschaft mbH (SRFG), Salzburg, Austria; project coordinator: Michael Thelen Project partners Independent L. ONLUS (IND), Italy Sustainability InnoCenter (SIC), Sweden VIABIRDS Technologies GmbH (VIA), Austria Universität Leipzig (ULEI), Germany Südtiroler Transportstrukturen AG – Green Mobility Department (STA), Italy Funding DUT Call 2022 – European Commission under the Horizon Europe Partnership scheme Funding is being provided by the Austrian Research Promotion Agency (FFG), Ministry of Enterprises and Made in Italy (MIMIT), the the Federal Ministry of Research, Technology and Space in Germany (BMFTR), and the Swedish funding agency (Vinnova) Project nr. F-DUT-2022-0007 Duration 01/2024 – 06/2026 Website https://sucolo.eu/ Deliverable nr. D2.3 Deliverable title AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations Authors Benjamin Gaunitz, Silvia Torres Landaverde, Viola Süß (ULEI) Version & status Version 2.0 Date 10.10.2025 D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 6 Purpose of the document This document explains the findings from Task 2.3 with the aim to create an AI-based simulation model for the optimal placement of micro-hubs and cargo bike pick-up stations as a case study in the city of Leipzig, Germany. The objective of this task is to use available open data (see data catalogue from Deliverable 2.2) to develop this model and simulate the best possible location for micro-hubs and stationary cargo bike pick-up stations. Regarding the challenges delineated in Deliverable 2.1 Review of challenges for sustainable goods logistics and delivery solutions in urban outskirts, the proposed locations consider inclusivity, barrierfree and social aspects. Although the simulation model is trained primarily from data from Leipzig, it would still be able to be extended for use in municipalities outside of these geographic areas. Additionally, in Task 2.4, a software prototype for bike couriers for delivery scheduling and routing is developed and can be used as a supplement to a holistic logistics concept for outskirts. The model will be further refined and developed during the pilot project phase. A further city, Merano, Italy is included in the model. Executive Summary This deliverable outlines the background and development of a simulation model that identifies optimal locations for pick-up stations for parcels, cargo bike rentals and micro-hubs in suburban or outskirt areas. These aforementioned stations are designed to improve the efficiency of last-mile logistics by integrating community aspects and delivery demands. The tool primarily targets urban delivery companies, promoting not only operational efficiency, but also the context-specific needs of outskirt residents. The simulation model takes the form as an app and is based on findings from a citizen survey in the Lützschena-Stahmeln district in Leipzig, Germany conducted in 2025. Resulting from this, three personas and user stories channeled the real-world requirements to the model. Interim results show that further refining of the model is necessitated, and further iterative developments will be included in D4.3 Reports on the research pilots’ design, implementation and results. In the end, the simulation frontend will be able to be found open access under the following link: https://github.com/Logistics-Living-Lab/sucolo-simulation-frontend D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 7 1. Motivation and background There is a lack of shops and facilities for everyday goods and services on the outskirts of Leipzig (1). Especially older people often rely on cars for their daily tasks, as walking distances can be difficult for them. In contrast, many points of interest (POIs) are located in the city center, which makes driving essential for people living on the outskirts. The SuCoLo project aims to provide more eco-friendly mobility, shopping, and delivery processes for these areas in order to connect them towards the 15-minute city ideal. Therefore, the development of simulation approaches is the subject of case studies in Leipzig, Germany and Merano, Italy. For the city of Leipzig, a new delivery concept for goods for the outskirts is developed. The concept involves home delivery via a cargo bike courier, utilizing a mobile micro-hub for lastmile parcel delivery. However, the model can also identify suitable bikesharing locations and parcel pick-up stations. The pick-up station location is based on POIs and detailed population data of the citizens in the area. The simulation model contains dynamic, social factors and preferences of the residents, such as barrier-free access to facilities, average age structure, average income, etc. In Merano, there are currently no decentralized self-service bicycle rental systems that would allow residents and tourists to easily use bicycles or cargo bikes for short trips. The simulation model can determine suitable locations for decentralized, station-based rental systems, considering geographic and urban characteristics. The SuCoLo Data catalogue of suitable and available (local) data sources/data sets (Deliverable 2.2) forms the database for this task (2). Open data from the municipal platforms of Leipzig and Merano are used to identify available population data. In addition, geodata from Open Street Map (OSM) is used to identify potential POIs. Based on this, the simulation model indicates the best possible location for micro-hubs or selfservice bike pick-up stations in urban environments and outskirts. This is presented as a heatmap, since multiple locations may be suitable under different circumstances. To evaluate the model, the cities of Leipzig and Merano are chosen as case studies in SuCoLo. 2. Problem description and model formulation Urban outskirts are generally not very well provided with infrastructure for everyday life as they lack fixed service facilities. As a result, people living in these suburbs have different needs from those living in the city center. Mobility, shopping, and consumer behavior require either travel or ordering online. These areas are generally sparsely populated, often with an older population and fewer commercial establishments. This is why people who live on the outskirts of the city produce significantly more emissions. Most of the pick-up stations, especially parcel lockers, are located strategically within walking distance of residential areas, which makes them easily accessible and encourages their use. They are usually located near public transport stations, business centers, financial areas, workplaces, gas stations, shopping stores, or cultural centers. Any place where there is a high concentration of people with high internet shopping frequency is attractive. These conditions are met by densely populated, inner-city areas. D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 8 For this reason, an AI-based simulation model has been developed that can consider social factors and the needs of residents on the outskirts. This allows for more sustainable, inclusive and efficient delivery of goods outside the city center. The model supports determining a suitable location for an inclusive pick-up station or a bike rental station. 2.1. Model assumptions, data and parameters for Leipzig There is a growing demand for last-mile delivery solutions, especially in low-density outskirts. Efficiently placing pick-up stations in these areas requires consideration of logistics performance, social accessibility, and demographic data. The model uses logistic regression and location knowledge to recommend high-potential pick-up sites. It evaluates infrastructure, accessibility, community needs and delivery history to suggest adaptable station placements that are optimized to cater to stated preferences (see below). Citizen Personas To identify potential user groups that would use the developed solutions, three different citizen personas were generated that would depict a typical citizen living in that area along with their wants, needs, values and fears. These personas are based on a citizen survey in 2025 in Lützschena-Stahmeln, Leipzig, Germany (n = 158). The three target groups have been identified include an elderly retired person, a family-oriented person who works from home and lives with his family, and a young student who prefers sports and entertainment. Figure 1 Three personas based on target groups These findings can be transferred to the simulation model app as ready-made user profiles. According to their assumed requirements, features, demographics and preferences, differing weightings are transferred across the three different scenarios. Each user profile or persona then stands for a possible target group that can use the pick-up stations (table 1). D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 9 Table 1 Example of Digital Personas for Model Use Persona Demographic Preferences and Weightings (example) Implication Persona 01 Older person, possibly retired or has limited mobility • "cafe"= 6 • "cafe_wheelchair"= 8 • "community_centre"= 8 • "community_centre_wheelchair"= 10 • "education_wheelchair"= 6 • "entertainment"= 8 • "entertainment_wheelchair"= 10 • "healthcare"= 6 • "healthcare_wheelchair"= 8 • "hospital"= 4 • "hospital_wheelchair"= 6 • "library"= 10 • "library_wheelchair"= 10 • "local_business"= 8 • "local_business_wheelchair"= 10 • "marketplace"= 10 • "marketplace_wheelchair"= 10 • "parcel_locker "= 5 • "parcel_locker_wheelchair"= 5 • "post_box"= 10 • "post_box_wheelchair"= 10 • "post_office"= 10 • "post_office_wheelchair"= 10 • "restaurant"= 8 • "restaurant_wheelchair"= 10 • "station"= 10 • "station_wheelchair"= 10 • "supermarket"= 8 • "supermarket_wheelchair"= 10 A socially active, accessibility-sensitive user. Pick-up stations in the neighborhood should be highly accessible and integrated into community-oriented environments. Persona 02 Middle-aged family-oriented person, employed • "gas_station"= 10 • "parcel_locker"= 10 • "post_office"= 10 • "restaurant"=10 • "bicycle_rental"= 4 • "post_box"= 8 • "station"= 6 • "shopping_centre"= 10 • "local_business"= 10 • "rental_service"= 4 • "marketplace"= 6 • "supermarket"= 10 • "entertainment"= 10 • "kiosk"= 6 • "cafe"= 10 This user values logistical efficiency. Ideal pick-up station locations should minimize detours, be close to transport infrastructure, and facilitate quick parcel access. D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 16 Figure 3 Simulation model framework 4. User interface of the “Location Finder App” The graphical user interface (GUI) of the simulation model, termed “Location Finder App” is divided into two parts. On the left side is the control panel and on the right side is the map. Figure 4 Screenshot of the app’s start screen D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 17 The user can select features (e.g., proximity to a post office), assign distances and optional penalties, and adjust the weighting sliders from -10 to +10 based on importance given to the listed features. This which directly affects the model output. Figure 5 Demonstration of the simulation model results based on set preferences D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 18 Table 5 GUI functions Screenshot Function Description Location Finder App App Name Logo SuCoLo Logo; Leads to https://sucolo.eu/ Radio Button City Selection By selecting a city, the map jumps to the chosen city. Resolution properties There can be three different resolutions be selected from 300m to 4.8km. Drop Down Menu Type of Feature For selecting a Type of Feature, which can be Nearest Amenity, Amenity Count, Amenity Present and District Feature. Checkbox Wheelchair Accessible Only shown after choosing the Type of Feature. Nearest Amenity Menu Only shown after choosing the “Nearest Amenity”. Choose Amenity. Options: select "Wheelchair Accessible” for barrier-free features, select a distance larger than 200 meters, and select a penalty for influencing the model. Amenity Count Menu Only shown after choosing the “Amenity Count”. Choose Amenity. Options: select "Wheelchair Accessible” for barrier-free features and select a distance larger than 200 meters. D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 19 Amenity Present Menu Only shown after choosing the “Amenity Present”. Choose Amenity. Options: select "Wheelchair Accessible” for barrier-free features and select a distance larger than 200 meters. District Feature Menu Only shown after choosing the “District Feature”. Choose a feature. Drop Down Menu “Select Amenity” for Nearest Amenity, Amenity Count and Amenity Present For adding one amenity to the model. Drop Down Menu “Select District Feature” for District Feature For adding one district feature to the model. D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 20 Apply Button By pressing the button, the Type of Feature, Amenity, Distance and Penalty can be chosen. Weight Slider; Delete the Selection The menu opens after pressing the Apply button. Users assign weights to features, adjusting importance and influence dynamically. Build Model Button By pressing the button, the calculations for the model will start. Model is shown on the map The model is built. Popup by clicking on a Hexagon Shows the calculated model scores. Reset Button Reload the page. Scenario Buttons Reloads predefined Scenarios (Personas). D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 21 Legend Score Scale Shows Score Scale of the model. Position on the bottom of the page. Map Zoom in (+) or out (-) Zoom function for the map, Position on top of the page. 4.1. Model Evaluation and further implementation The model is assessed through a simulation case study and user persona-driven profiles. 4.2. Case Study in Germany, Leipzig The model was tested using open data from Leipzig, particularly the Lützschena-Stahmeln district. As stated earlier, citizen surveys were used to build user personas, incorporating their amenity preferences and demographic profiles. Findings show that the model can adaptively suggest pick-up points in underserved suburban areas with both logistical and social benefits. Stations also serve as community hubs, encouraging interaction among residents and couriers. Figure 6 Screenshot of the app centred in Leipzig D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 22 5. Discussion and further work The simulation model developed for the optimal placement of micro-hub and cargo bike pickup stations on urban outskirts demonstrates the potential of combining data-driven logistics with community-focused urban planning. By integrating open spatial data, demographic profiles, and delivery analytics, the model offers a dynamic and adaptable solution fitting to the unique characteristics of less densely populated areas. The case study in Leipzig illustrates how such a model can be effectively applied in real-world settings. Incorporating user personas derived from citizen surveys enables the system to align logistical priorities with local social preferences, ensuring that proposed locations are not only operationally efficient but also socially inclusive. This dual focus enhances the model's relevance for stakeholders aiming to balance economic goals with community engagement. One of the model’s uniqueness is its interactivity and flexibility. Users can adjust weights and parameters to simulate various scenarios and business objectives. The ability to connect realtime data sources and user feedback ensures that the model remains responsive to changes in urban dynamics, such as population shifts or evolving delivery patterns. However, some challenges remain. The quality and granularity of open data sources can affect model accuracy. As said, available open data sources for Leipzig and Merano were used. There are some gaps in the historical data, which have been filled in with supplementary data. Further data availability, such as cycle paths, population figures in smaller units, etc., would significantly enrich the model. Moreover, it has been found that social aspects are harder to quantify and integrate systematically. Future development could incorporate additional machine learning techniques for more advanced prediction, deeper integration of behavioral data, and broader testing across different cities to enhance generalizability. Future developments to the simulation model will be outlined in D4.3 Reports on the research pilots’ design, implementation and results. More potential features will be possibly added to the app in the future, including: • Integrating other European cities • Predictive modeling for future delivery demand on the outskirts • Automated ETL pipeline implementation with real-time data • Allow users to visualize suggested locations on a city map and use it for model calculation • Integrate co-creation features, such as user feedback on location suggestions, integrating occasional community events, etc. • Highlight locations with potential for social interaction (e.g., parks, cafés, local markets) In summary, this model provides a new tool for modern urban logistics planning, especially in the context of growing demand for sustainable last-mile delivery solutions. By combining technical precision with social awareness, it sets a promising direction for the future of mobility and urban service infrastructure. In the end, the simulation frontend will be able to be found open access under the following link: https://github.com/Logistics-Living-Lab/sucolo-simulation-frontend D2.3 AI-based simulation model for optimal placement of micro-hubs and cargo bike pick-up stations SuCoLo 23 References (1) Stadt Leipzig. Amt für Statistik und Wahlen Leipzig. (2023). Kommunale Bürgerumfrage 2023. Retrieved April 24, 2025, from https://static.leipzig.de/fileadmin/mediendatenbank/leipzigde/Stadt/02.1_Dez1_Allgemeine_Verwaltung/12_Statistik_und_Wahlen/Stadtforschung/K ommunale_B%C3%BCrgerumfrage_2023_-_Ergebnisbericht.pdf (2) LLL GitHub (n.d.). SuCoLo DataCatalogue. Retrieved April 24, 2025, from https://github.com/Logistics-Living-Lab/SuCoLo_DataCatalogue (3) Stadt Leipzig. (n.d.-a). Ortsteile und Stadtbezirke - Leipzig-Informationssystem. Retrieved July 10, 2024, from https://statistik.leipzig.de/statdist/index.aspx (4) Stadt Leipzig. (n.d.-b). Informationen zum Leipzig-Informationssystem. Retrieved July 10, 2024, from https://statistik.leipzig.de/statserv/servinfo.aspx (5) Stadt Leipzig. (n.d.-c). Open Data-Portal der Stadt Leipzig. Retrieved July 2, 2024, from https://opendata.leipzig.de/ (6) OpenStreetMap contributors. (n.d.). OpenStreetMap. Retrieved July 2, 2024, from https://www.openstreetmap.org/ (7) Gemeindeverordnung über die Stadtviertel (2023). Gemeinde Meran. Retrieved October 10, 2025, from https://www.gemeinde.meran.bz.it/de/Verwaltung/Dokumente/Gemeindeverordnung_ueb er_die_Stadtviertel. (8) Data browser (n.d.). ASTA data. Retrieved October 10, 2025, from https://statastat.prov.bz.it/databrowser/#/de. (9) Open Data Portal Bozen (n.d.). CIVIS. Retrieved October 10, 2025, from https://data.civis.bz.it/de/.