I4C – Intelligence for Cities:
KI-basierte Anpassung von Städten an den Klimawandel – von Daten über Prädiktion zu Entscheidungen
Contact
- Prof. Dr. Bernhard Neumärker
bernhard.neumaerker@vwl.uni-freiburg.de - Julian Hübner (responsible)
julian.huebner@vwl.uni-freiburg.de
Thematic Overview
In the I4C project, digital innovations in the form of artificial intelligence (AI) methods are developed and used to adapt cities to the challenges of the climate change. A process chain from data acquisition to analysis and environmental prediction to concrete measures is being tested. The project is accompanied by considerations of ethics and the reciprocal transfer of knowledge when dealing with AI. The results of the project team anchored in the Freiburg region will be demonstrated within the Green City Freiburg.
Today, more than 75% of the German population lives in cities. A large proportion of the gross domestic product is generated there. Cities are the cradle of life and trade. Extreme weather events such as floods, heat waves or storms, which will occur more frequently due to climate change, are increasingly inducing challenges for cities. Adaptation to the changed conditions is essential for the long-term protection of our population and economic viability. Digital innovations can make an enormous contribution to capacity building for climate adaptation. Due to the complexity of the city systems, the methods of artificial intelligence (AI) play a special role. With AI, complex calculations such as short-term and local forecasts of extreme events, long-term projections of risks as a basis for planning or intelligent real-time control can be implemented efficiently and reliably.
I4C tackles these challenges. An AI-supported process chain will be set up, which will ultimately deliver measures to improve the adaptability of cities to extreme weather events. Meteorological process models and climate simulations are used to identify areas in cities that are susceptible to thermal stress, flooding and storm damage. The stress situations and the associated effects are simulated, quantified and visualized in a semantic, building-resolving 3D model. Based on the simulations, experts propose concrete measures that can be systematically applied to politics, law and planning. Control elements, e.g. building automation, will be automatically derived. Ethical considerations for implementing AI and reciprocal knowledge transfer build the foundation for I4C.
AI methods, especially deep learning and predictive analysis, are used in all project phases: environmental data is automatically segmented semantically and projected into the 3D model of a city; Sensor networks are evaluated and controlled with deep neural networks, which improves sensitivity; Simulation results, together with weather data, climate projections and the 3D model, form the data basis for real-time environmental prediction on small scales (down to meters); Buildings are intelligently controlled based on prediction using new deep reinforcement learning methods and optimized for minimum consumption and high comfort; Possible causes of undesirable effects are identified with the help of backtracking in the predictive neural networks and help scientists and local decision-makers to investigate and implement digital innovations and reforms in a targeted manner. I4C will thus produce numerous innovations both on the technical side of the AI and in the application.
The process chain is being tested using the city of Freiburg as an example. To realize I4C, an interdisciplinary Freiburg team from the Albert Ludwig University and Fraunhofer institutes in the fields of computer science, engineering, meteorology, hydrology, economics, politics, law and social sciences is joining forces with companies from the region from the field of mobile mapping , sensor production, urban planning, energy supply and consulting, local authorities and the Green City Freiburg.
The Götz Werner chair deals with economic policy issues relating to acceptance, regulation, reform conditions and implementation options for AI-based applications in climate adaptation strategies.
Projects
- TBA