Building dynamic spatial information for live digital twins in urban areas
▲ Research Fellow Yoon Joon-hee and Research Specialist Kim Ji-eun, Department of Future & Smart Construction Research, KICT
From Static to Live Digital Twin
Land Digital Twin technology is evolving from Static Digital Twin into Live Digital Twin. In 2003, Dr. Michael Grieves proposed the concept of the "Digital Twin," based on the idea that interaction can be established and intelligence achieved through twinning (or mirroring) of the physical and virtual worlds. Subsequently, with advancements in data transmission, visualization, and platform technologies, digital twin technologies and their platform development have progressed across various domains. From a construction and land management perspective, digital twins can be viewed as a technology that converges, interprets, and visualizes the "shape information" of structures such as buildings, roads, and terrain with "phenomenon information" like population movement, traffic flow, weather changes, and infrastructure transformations, to solve diverse social challenges. Until now, Land Digital Twin has primarily focused on analyzing and visualizing shape information or making it into platforms, consequently limiting its scope of analysis. Now is the time to focus on phenomenon information. A true Live Digital Twin will be completed by combining near real-time phenomenon information with the shape information platform of Static Digital Twin.
Dynamic Spatial Information for Live Digital Twin
For a Live Digital Twin, the construction of dynamic spatial information is essential. To store, extract, visualize, and analyze information in a digital twin platform, the location and attributes of each piece of information must be acquired and stored. Information with assigned locations and attributes is called spatial information. From the perspective of digital twin development, spatial information needs to be classified into static and dynamic spatial information from the viewpoints of shape and phenomenon.
If static spatial information is spatial information that is consistent over the long term, like buildings and roads, dynamic spatial information can be defined as spatial information (Dynamic or Temporary Spatial Information) that exists temporarily from an SOC perspective, such as pedestrians, vehicles, and facility damage, which changes or disappears. Static spatial information can be updated on a cycle of several days to several months. It is generated by the Ministry of Land, Infrastructure, and Transport and local governments in accordance with laws, and also is independently generated and used by companies like Google, Naver, and Kakao. On the other hand, dynamic spatial information has an update cycle of several minutes to several days. Currently, the spatial and object targets of dynamic spatial information are vehicles on major roads, with information provided and updated through CCTV, probe cars, and driver reports, which have limitations in terms of their update cycles and spatial recognition range. However, recent developments in AI-based image processing, IoT, drones, Urban Air Mobility (UAM), and satellite technologies are making it possible to overcome such limitations.
KICT's Dynamic Spatial Information Construction Technology
The Korea Institute of Civil Engineering and Building Technology (KICT) has been leading a project titled "Development of Dynamic Thematic Map Construction Technology Based on Fixed/Mobile Platforms for Next-Generation Digital Land Information," with a total budget of KRW 18.2 billion, since 2022. This project is one of the four core initiatives of the Digital Land Information Technology Development Program, managed by the Korea Agency for Infrastructure Technology Advancement (KAIA). The KICT has been developing technologies to generate and update dynamic spatial information in near real-time and represent it accurately. This project defines dynamic spatial information as information occurring in urban living SOC, including moving objects and changing phenomena. Its goal is to develop dynamic information thematic map construction technologies through continuous near real-time detection and tracking using fixed sensors (CCTV, Wi-Fi, etc.) and mobile sensors (drone stations) to solve various social problems. While CCTV allows 24-hour monitoring but has a limited area of coverage, drones (stations) can cover wider areas but cannot monitor 24/7. This project aims to merge the advantages of both platforms for urban area monitoring. Figure 1 illustrates the concept of urban monitoring based on ground-fixed and airborne mobile sensors. The project consists of three main core technologies: "Development of Dynamic Information Collection Technology Based on Fixed Platforms," "Development of Dynamic Information Collection Technology Based on Mobile Platforms," and "Development of Dynamic Information Analysis, Prediction, and Representation Technology." These are further divided into a total of six core sub-technologies, as shown in Figure 2.
In "Development of Dynamic Information Collection Technology Based on Fixed Platforms," specifically within the "Development of Heterogeneous Sensor Linkage and Mobile Information Collection Technology" section, the research focuses on the real-time detection and tracking of mobile objects, which is achieved by integrating object detection and tracking technologies with fixed sensor equipment such as CCTV, Wi-Fi, and Bluetooth. This involves analyzing the environment of fixed platforms and developing methods for acquiring and collecting sensor information, creating data models for transmitting and storing mobile object location data using heterogeneous sensor data, interconnecting different sensors, recognizing and classifying mobile objects, and extracting the location information of mobile objects in heterogeneous sensing environments. In the "Development of Continuous Time-Series Mobile Object Information Tracking Technology Based on Fixed Platform Linkage" section, the research advances the development of mobile object data models for continuous location tracking. This is achieved by collaborating with fixed sensor equipment to monitor specific areas and developing technologies for seamless location handover between homogeneous and heterogeneous sensors. The goal is to enable the continuous tracking of mobile objects' time-series location data within urban environments.
In the part titled "Development of Technology for Collecting Information Using Fixed Platform Heterogeneous Sensor Integration and Mobile Object Data," the research discusses how sensor devices that are fixed in place, such as CCTV, Wi-Fi, and Bluetooth, can be utilized to detect and track objects in real time using object detection and tracking technologies. The research includes analyzing fixed platform environments, developing methods for acquiring and collecting sensor data, and creating models for transmitting and storing mobile object location data using heterogeneous sensor data. Additionally, the project aims to develop technologies for the recognition and classification of mobile objects using heterogeneous sensors, as well as for extracting the location information of mobile objects in heterogeneous sensing environments.
The "Development of Dynamic Information Collection and AI Learning Data Construction Technology" section focuses on collecting dynamic information in urban areas and constructing AI learning datasets. To achieve this, drone/mobile platforms and operation systems for dynamic information collection are established, taking into account the characteristics of each test bed region. In addition, technologies for converting learning data, automatic classification, and the automated construction of multi-dimensional dynamic information datasets are developed by topic.
In the "Development of Knowledge/Learning-based Dynamic Information Recognition Technology" section, the research aims to develop knowledge and learning-based dynamic information recognition and integration algorithms using data collected from mobile platforms. The goal is to enable collaborative and continuous object recognition between fixed and mobile platforms. The research involves developing dynamic information data integration algorithms that can account for spatio-temporal changes, as well as technologies for object-specific dynamic information recognition, classification, and situation detection. Furthermore, the research is focused on creating collaborative object observation technologies between fixed and mobile platforms to visualize the outcomes of these efforts.
Finally, the "Development of Dynamic Information Analysis, Prediction, and Representation Technology" part analyzes the results from the previous two parts and constructs dynamic thematic maps based on these findings. The "Development of Dynamic Information Analysis and Prediction Technology Based on Movement Context Information" section aims to generate movement context information by linking object-level movement information and static data collected from fixed/mobile platforms, with the goal of applying AI to movement analysis and prediction. To achieve this, the research is developing technologies for: linking static data and data mining for movement context information generation, creating movement time-series pattern information and context information, and applying AI technologies for movement analysis and prediction based on context information.
In the final "Development of Dynamic Thematic Map Construction and Update Technology" section, utilizing the previously developed dynamic information, the research seeks to construct and update user-customized dynamic thematic maps. This involves identifying dynamic thematic map service models from public and private sector perspectives, developing 2D/3D visualization technologies for multi-dimensional dynamic information including location, time, and status, and creating technologies for user-customized dynamic thematic map construction and updates. Figure 3 illustrates an example of a dynamic thematic map. The project has in particular focused on establishing early test beds from the first year to successfully demonstrate the project, verifying annual achievements. Leveraging the existing experimental infrastructure from previously completed intelligent crime prevention and immersive disaster research units at the KICT, the project aims to minimize research and development risks by working in close collaboration with Anyang City as a local government demonstration site. This includes establishing drone/operation platforms within the Anyang City test bed and acquiring actual urban data such as CCTV footage and IoT sensing data. In addition, for dynamic thematic maps, the project is identifying and implementing user-oriented demand-based dynamic thematic maps through regular commercialization consultation meetings involving key public institutions, local governments, and private sector stakeholders.
