Milan’s digital twin project

The Italian city of Milan is pioneering a new era of urban management with its comprehensive digital twin of the metropolitan area. It was developed in a municipal project involving four leading geoinformation companies.

Imagine a city that can be seen, measured and managed from the office rather than through burdensome and time-consuming on-site visits. A city that can respond to its inhabitants’ needs while optimizing its resources. A city that can anticipate and solve problems while creating new opportunities for innovation and collaboration. This is not a fantasy, but a reality. City digital twins are shaping the urban environment, creating a dynamic and holistic model. The Italian city of Milan, a global hub of commerce, fashion, finance, research and tourism, has embraced this vision, resulting in a comprehensive digital twin of the metropolitan area.

The project to create a digital twin of Milan involved the Municipality of Milan and four leading geoinformation companies: Compagnia Generale Ripreseaeree (CGR SpA) for aerial survey, CycloMedia for mobile mapping, Esri Italia SpA for data management, and Servizi Informazione Territoriale (SIT Srl) for topographic survey. The Laboratory of Geomatics at the University of Pavia also participated in the data quality analysis and assessment. The resulting digital twin offers a rich and diverse dataset, capturing various aspects of the urban environment from different perspectives. It combines nadir and oblique aerial images, Lidar points and terrestrial mobile mapping data to create a dynamic and comprehensive city model. The project also delivers innovative products, such as true orthophotos (RGB and colour infrared [CIR] imagery), classified Lidar point clouds, DTM and DSM models, and a database of 22 urban objects such as traffic lights, outdoor seating, gates and road signs.

The project covered an area of approximately 1,575km², including 133 municipalities and the city of Milan. The aerial survey used the Leica City Mapper-2 hybrid sensor, which captured both nadir and oblique images and collected Lidar point clouds. The acquisition requirements specified that the aerial images must have a 5cm resolution and the point cloud density must be at least 20 points/m². The survey produced more than 434,000 images and over 220 billion points, requiring more than 50TB of storage space. The mobile mapping system (MMS) focused on Milan, where CycloMedia’s sensor covered almost 2,600km of roads. The MMS data had to have an image resolution of 8mm per pixel or less, and a point cloud density of at least 1,500 points/m² (both thresholds refer to data acquired at 10m from the vehicle path). The MMS data occupied more than 10TB of memory, including the imagery, point cloud and a database of about 1.2 million elements.

High-quality ground control network

One key component of the Milan digital twin project is the high-quality ground control network that ensures the accuracy and reliability of the data. The project team set up 200 control areas across the metropolitan area, each consisting of two markers about 100m apart. One marker is a stable topographic nail that serves as a geodetic reference, while the other is a photogrammetric marker painted on the ground. The flat area around the second marker allows for evaluating the Lidar vertical quality and density.

The Milan project partners opted for conventional static measurements to establish a trustworthy ground network for data quality assessment rather than using NRTK GNSS survey technology. This choice resulted in an average root mean square error (RMSE) of 8mm for the East, 9mm for the North, and 13mm for the Up direction. These results are well-suited for assessing the quality of the photogrammetric processing, given that the images have a ground sample distance of 5cm. Moreover, a terrestrial laser scanner was used to survey 50 out of 200 control areas to analyse the horizontal and vertical components of the Lidar data. Lastly, the project team used photogrammetry to collect 2,000 data points along the city’s main roads. These coordinates were compared with those obtained from the MMS survey to assess the accuracy and consistency of the MMS with the aerial dataset.

Comprehensive aerial and terrestrial surveys

Aerial data is another vital project component, with photogrammetric and Lidar data being captured to cover the whole metropolitan area. The quality of these datasets and the main aerial products was carefully assessed using various parameters and criteria. One of the innovative aspects of the project was the optimal management of overlaps. Modern cameras are distinguished by a wide variety of focal lengths and sensor sizes, which influence the field of view (FOV) and thus challenge the assumption that fixed percentages of overlaps are universally applicable. Photogrammetric blocks designed with fixed overlapping may then be more or less vulnerable to perspective obstructions connected to the FOV, and the phenomenon of ‘building leaning’ (or ‘apparent inclination of buildings’) must be considered.

Another interesting project result was the higher-than-expected point density of the Lidar data, which was around 55 points/m² instead of the required 20. This was due to the hybrid system allowing simultaneous imagery and points to be acquired. This higher value derives from the sensor’s characteristics (spiral-shaped scan pattern with a tilted axis) and the block design. The system enables the contextual acquisition of imagery and points; photogrammetric planning prevails on Lidar, resulting in increased point acquisition due to the imagery’s more considerable overlap.

The quality of the imagery, Lidar data and main aerial products was then assessed. The results show a high level of performance: the bundle block adjustment achieves an accuracy of 1.8cm and 3.7cm for the horizontal and vertical components, respectively. This corresponds to almost 0.5 and 0.75 of the ground sample distance (GSD). The vertical precision and accuracy of the Lidar data are 2.0cm and 4.8cm, respectively, while the horizontal precision and accuracy are 8.7cm and 12.2cm, respectively. Based on the quality of the acquired data, the products demonstrate an equivalent level of excellence. The orthophotos, both RGB and CIR, present a 2D error of 2.5cm, while the DSM and DTM have vertical accuracy of between 4 and 5cm. Lastly, the Lidar point cloud classification was thoroughly examined to ensure correct attribution to the 11 categories specified in the tender. These categories include underground, ground, electric lines, bridges, water, buildings, high vegetation, medium vegetation, low vegetation, outlier, and overground (meaning artificial objects not belonging to other categories).

Mobile mapping data similarly constitutes an essential element of Milan’s digital twin project. It facilitates the examination of the city in areas that pose challenges for traditional aerial surveys, particularly those characterized by narrow streets or dense vegetation, as commonly found in Italian urban areas. Moreover, mobile mapping is the primary data source for creating the urban object database. Therefore, a thorough evaluation is required concerning the resolution of panoramic images and the density of the generated points cloud. Both requirements have met the specified threshold of 8 mm per pixel for image resolution and 1500 points per square meter for point density (at 10 meters for the vehicle path).

One of the project’s core principles is to use both aerial and terrestrial data to gain complementary perspectives and improve the accuracy of our analyses. It is essential to ensure alignment between these datasets, and this alignment must be thoroughly assessed. A total of 2000 points were measured using photogrammetry and compared with the MMS data. The results show a satisfactory RMSE of approximately 7 cm horizontally and 14 cm vertically, indicating strong compatibility between the two data sources.

Software tools for accessing complex datasets

One of the key aspects of the tender’s technical specifications was the provision of advanced software tools for easy access to complex datasets. This is a significant improvement in optimizing urban data management and facilitating collaboration among local entities. Two main tools were proposed to achieve this goal. The first tool is a web app developed by CycloMedia that allows users to view, analyse and share geospatial information in a user-friendly way, which is helpful for urban planning and resource management. The second tool is a plugin for the Esri ArcGIS environment that enables the integration of complex datasets into existing workflows, thus improving the efficiency and accuracy of geospatial analysis. These innovative solutions will allow the Municipality of Milan to make well-informed and timely decisions based on accurate and up-to-date data. In short, implementing these software tools marks a crucial step in urban data management, fostering more collaboration and information sharing among various local departments. By adopting these technologies, Milan will be better prepared to face future challenges and improve the quality of life for its citizens.

A catalyst for social change

Milan’s digital twin project represents a state-of-the-art example of how geospatial data can create a realistic and dynamic representation of a city, enabling better governance, planning and innovation. By integrating different types of data from aerial and terrestrial surveys, the project provides a comprehensive and accurate view of the urban environment, its features and its changes over time. The project also demonstrates how advanced software tools can facilitate access to and analysis and sharing of complex datasets, enhancing collaboration and decision-making among various stakeholders.

The project’s outcomes will benefit the Municipality of Milan and the citizens, businesses, researchers and visitors who live, work or interact with the city. The project is a technical achievement and a model for the future. It shows how a digital city can offer new possibilities for improving the quality of life, the sustainability and the resilience of urban areas. It also shows how a digital city can foster creativity, experimentation and participation, stimulating new ideas and solutions for the common good. The project invites us to imagine a world where every city has a digital twin, where data becomes a valuable asset for enhancing urban development, and where technology is a catalyst for social change. The project challenges us to embrace this vision and make it a reality.