This workgroup focuses on detection, segmentation, and contextualization of entities related to the smart building domain. The aim is to discover and learn the relationships between actors and objects within a common spatial reference frame to enhance the capabilities of self-aware buildings. With state of the art technologies, methods and concepts for knowledge integration in structural environments are researched. Therefore, the workgroup „Cognitive Computing for Self-Aware Buildings“ unites competencies from various disciplines such as Computer Vision, Data Science, Civil Engineering, and Machine Learning to path the way to Cyber Intelligent Buildings.
Sensor data fusion in the intelligent building
The research area sensor data fusion is based on raw data that are gathered by multiple different sensors in an intelligent building. It takes a big amount of data as given, so the research topic is to interpret them in an intelligent way and draw conclusions with for example artificial intelligence.
One approach is to analyze the data with deep learning algorithms and get to know how many people occupy a room. Therefore different combinations of sensors will be tested, to compare the quality of information they offer. If the accuracy of occupancy estimation is high enough, and the system has enough test-data with known occupancy-count, the next step is occupancy-prediction. If a system is able to predict how many people will be in a room in one hour, it can adjust the heating or cooling before the temperature becomes uncomfortable, not when sensors show that the air is already inconvenient.
The necessary information will be gathered by pre-installed sensors as part of a new intelligent building on the one side, and by sensors composed and placed for this research particularly on the other side. One more task of this research area is to build sensors useful for research area one and research area three of this institute.
BIMiB Structural System
Contrary to newly built structural environments, the data on pre-existing buildings intended for modification is rarely extensive enough for direct economic use in modeling. For this reason, a system is being developed which reconstructs a digital Building Information Model from 3D laserscan point clouds and digitized drawings of the structure. New approaches to ontology-based semantic knowledge representations in combination with machine learning and pattern recognition are being researched for their combined potential to increase automatic detection rates.
DORIOT | Dynamic runtime for organically (dis-)aggregating IoT-processes
The Institute for the Intelligent Building is researching aspects on the Internet of Things (IoT) within the context of SmartX environments. SmartX environments serve users with multiple services, which depend on capabilities of available devices within a network. Requirements, availability and capabilities of each device might change over time, e.g. due to software-updates, device configurations or technical issues. In order to improve the quality of services, it is required to dynamically aggregate or disaggregate processes with respect to changes in SmartX environments. Therefore, a dynamic runtime environment is required to control (dis-)aggregation of processes in the Internet of Things. The Institute for the Intelligent Building focuses on verifying and validating current developments from partners found in research and industry for the following working areas: IoT runtime environments, optimization of aggregation and disaggregation via organic computing, and IoT security and IoT safety.