The FlatCity project imbricates into the smart city the knowledge of the barriers that affect the mobility of its inhabitants. The technology developed in the project will capture data from different sources and generate algorithms to transform that data into information on urban elements that affect the user mobility in public spaces (streets and buildings) by obtaining data from open sources available on the Internet (eg, schedules of public transport or incidents in the same, works in the city or events of massive interest), through geomatic sensors (such as LiDAR sensors), and through crowd sensing techniques using both hardware and software sensors of the users of the city, in particular those included in mobile or wearable devices carried by the citizens (eg, accelerometers, gyroscopes, barometers or light and sound sensors) or in the marks that citizens leave on the Internet as in social networks.



+ info



+ info

Velocity: Procesado eficiente de Big Data espacio-temporal para FlatCity


+ info


The AUDACity (Analytics Using sensor DAta for FLATCity) subproject is integrated within the FLATCity coordinated project through the automatic detection of barriers that affect the mobility of citizens. Those barriers will be automatically detected through both SW sensors and HW sensors carried by the users themselves (automated learning through crowd sensing from mobile and wearable sensors, as well as traces that citizens leave on the Internet). The extracted information, which will be valuable for the identification of accessible places and routes, will be stored in the spatio-temporal database of FLATCity.

Regarding the data obtained from SW sensors, AUDACity focuses on the extraction of knowledge from open data, coming both from public services and private service providers. There will be two separate data acquisition mechanisms. On the one hand, AUDACity will extract the accessibility information that some open data sources such as OpenStreetMap provide. On the other hand, AUDACity will continuously monitor social networks. In particular, the system will collect contents by following people and organizations identified as relevant in the area of disabilities and accessibility, as well as keywords that are frequently used in this area. Information extraction and statistical classification techniques, with which the UC3M team has experience, will be applied to those contents in order to detect places in which there might be obstacles or other situations that hinder mobility.

Regarding the data obtained from HW sensors, AUDACity will detect physical barriers (stairways, ramps, zones with mobility difficulties) from accelerometers, pressure sensors and physiologic sensors (e.g. heart rate and skin conductivity sensors). Both wearable sensors and sensors embedded in mobile devices will be used. The detected barriers will be identified, geolocated and automatically annotated with mobility-related information. The evolution in time of those obstacles will be tracked. Each physical barrier will also be associated with a difficulty index, which estimates the effort the obstacle supposes to each user. Machine learning techniques will be developed in order to compute those indexes from the datanthat sensors provide. The cameras and microphones of mobile devices will also be used in order to identify other barriers affecting visual or hearing impaired people. Finally, minimally-invasive mobile user interfaces will be developed in order to make the capturing of information easier for the end-user.

The technology developed by AUDACity will be adapted and optimized for its deployment in the distributed computation systems developed at the level of the FLATCity coordinated project. Its integration in the common prototypes of FLATCity will maximize the results and impact of the research activities.


In the recent years, the reconstruction of three-dimensional models of cities and buildings has emerged as a research topic of great interest due to the growing demand of updated and detailed models. The reconstruction of digital 3D urban environments is essential for many applications for which traditional 2D cartographic models are not enough. Within the broad set of applications requiring 3D data, path planning and accessibility diagnosis arise as applications for which a representation of the as-built environment from a three-dimensional perspective is essential for realistic analysis.

Laser scanning is a consolidated technology for the collection and analysis of three-dimensional data on the as-built status of large-scale civil infrastructures. However, point clouds are composed of massive and raw information that should be processed to extract the information that is useful for the applications they are intended to serve. Therefore, it is necessary the development of algorithms for the automatic processing of point clouds.

The main objective of the subproject RunCity (3D route modelling for FLATCity) is the automatic generation of updated and detailed 3D cartography of urban environments from point clouds. The challenges of the subproject are, on one hand, the development of methodologies for the automatic processing of urban point clouds, and on the other hand, the extraction of parameters of interest for the accessibility diagnosis. Special attention will be given to elements composing the navigable space for pedestrians such as ramps, stairs, transit areas and obstacles to navigation. In addition to geometric information, methodologies will pursue the extraction of topologic and semantic information in order to generate coherent graphs for path planning.


Managing massive volumes of spatial and spatio-temporal information poses problems to current database management systems. First, the information is very diverse: point clouds collected by LiDAR sensors, raster data collected by complementary sensors, vector information collected by hardware sensors carried by users, and semi-structured and unstructured information collected from software sensors and open data sources. Second, the volume of information is large (for example, the size of a LiDAR point cloud can easily reach several gigabytes). And thirdly, response times to user queries must be small since it is not feasible that a user requiring a route would have to wait a large response time. These three features not only complicate the storage of information but also make the definition of information processing algorithms more complex efficient. Although the first steps have been taken in the development of Big Data technologies for the storage and processing of geographic information, there is a great need for research in this field: current solutions consider only one of the logical models, do not allow queries in which information from various models is used, they focus more on performing batch tasks than in performing real-time tasks, and they are based on the use of distributed versions of traditional structures instead of defining new structures that take advantage of distributed processing features.

The main objective of the subproject VeloCity: Efficient processing of spatio-temporal Big Data for FLATCity is the development of algorithms, data structures and storage and information processing technology that can be used for the efficient extraction of semantic models from massive geographic information obtained from mobile sensors, fixed sensors, open data and data generated by volunteers. This main objective is broken down into the following specific objectives:

1) Definition and implementation of data structures for compact storage of spatio-temporal information and algorithms for efficient in-memory and distributed query processing.

2) Definition of LiDAR data processing algorithms adapted for its implementation Big Data using technologies.

3) Adaptation of processing algorithms for information collected by HW and SW sensors for its execution using Big Data technologies.