About

We, a team of researchers at ETH Zurich, Switzerland and Bauhaus­ University Weimar, Germany, applied Machine Learning methods to investigating how humans perceive their urban environment and what features of urban environment influence the humans' physiological arousal states. For this purpose, we conducted a study, where participants took a leisure walk in an urban environment. Empirical measures of participants' physiological conditions and urban environment features were analyzed using Machine Learning methods.

This research is a part of our research project "ESUM­-Analysing trade­offs between the energy and social performance of urban morphologies," at the Chair of Information Architecture, ETH Zurich, Zurich, Switzerland and the Chair of Computer Science for Architecture, Bauhaus ­University Weimar, Germany.

Commentary on ESUM (read) and Introductory audio slides:


Presentation and talks on ESUM:

ESUM complete picture (slides)
Machine learning and pattern analysis for ESUM (slides)
MAPS: Visual representation of results (slides)
Talk delivered in ITA, ETH Zurich, Switzerland (slides)
Falling Wall competition Paris, France (presentation won 3rd place): (slides)
Information Visualization: People's commuting experience in cities (slides)

Please cite this research work:

Ojha VK, Griego D, Kuliga S, Bielik M, Buš P, Schaeben C, Treyer L, Standfest M, Schneider S, König R, Donath D, Schmitt G (2018) Machine learning approaches to understand the influence of urban environments on human’s physiological response, Information Sciences, Elsevier (pdf).

ESUM, (2017) Human perception of urban environment, https://www.esum.arch.ethz.ch

ESUM

ESUM: Analysing trade-offs between Energy & Social potential of Urban Morphologies

In this research project, we investigate the impacts of urban morphology (UM) on citizen's social potential as a function of accessibility and perception and compare it with the impacts of UM on building energy consumption for parallel case studies in Zürich Switzerland and Weimar Germany. This is of particular interest since urban planning decisions of urban morphology have long–term implications that affect not only the energy performance of the building stock but also on people's experiential quality of the city environment.

In order to understand, measure and predict the impacts of UM on the perception of citizens we developed a three-tiered empirical study, which draws upon the participants' experiences in a real-world experiment, along with two controlled experiments using a 360­degree video and a gray-scaled 3D virtual reality (VR) model of the city. All three experiments are conducted using the same pre­defined path in each of the two cities, for a total of six participant based experiments. In each experiment, the participants subjectively rate the urban environment through surveys for an understanding of their explicit perception.

We also go beyond subjective impression ratings to gain further insight into the implicit perception of the participants by simultaneously measuring electro-dermal activity (EDA) data collected via wearable devices. The EDA data is processed to quantify the frequency, duration, and location of arousals along the experimental path to show a measure of excitation. The explicit perception data provides the sentiment of the experience such as spacious/narrow, beautiful/ugly, or private/public (among others) while the implicit perception provides the “raw” arousal-based perception from participants.

We also consider the impacts of urban environment features on perception, including temperature, illuminance, sound, and dust using data collected via a “sensor-backpack,” specifically designed for the use in the real-world participant studies. The data produced from this experiment is highly complex; the seven sensors have varying frequencies and varying temporal dependencies. We, therefore, developed an information fusion mechanism to combine the spatial-temporal data to produce a structured dataset, which is analyzed using four state-of-the-art Machine Learning algorithms. The quantified physiological responses or “human perception” data are paired with the sensor-based environmental measurements using Machine Learning based knowledge discovery approaches. These include classification, fuzzy rule-based inference, feature selection, and clustering to derive underlying relationships between participants’ physiological response and environmental conditions. The results indicate that the participant’s implicit responses are affected by both urban morphological features measured as participant field-of-view and street width as well as the dynamic environmental features temperature, noise levels, illuminance and traffic speed.

The subsequent experiments are further abstractions of the real-world experiment to emphasize and evaluation the impact potential of urban morphology. For example, the 360­degree video holds all dynamic environmental variables constant, and the gray-scaled 3D block geometry VR model purely focuses on participant response to the change in geometry, eliminating the influence of aesthetic qualities. The three experiments enable us to identify the “hot spots” of arousal as a function of pure urban form, thus gaining insight for how human perception is linked to UM.

For building energy demand and social accessibility potential, we draw upon existing methods to evaluate scenarios at the district scale (1km2) for the building stock where the participant experiments are conducted. The energy simulations utilize inputs that highlight the impacts of the pure urban form including the building volume, usable floor area and adjacency with neighboring buildings. Accessibility similarly uses urban morphological features as the primary input parameters to simulate select Isovist and centrality measures. The results of the energy and accessibility analysis are inherently linked to descriptors of UM and are compared to perception using a methodology that normalizes the data along the street network configuration. As an outcome, we develop a predictive model to estimate the effects of pure UM on perception and quantitatively compare this with the energy demand and accessibility potential of neighborhoods.


Partners


     

ETH Zürich, Chair of Information Architecture, Zürich, Switzerland


Principal Investigator:

Prof. Dr.-Ing. Gerhard Schmitt
ETH Zürich
Chair of Information Architecture
Switzerland



     

Bauhaus-University Weimar, Chair of Planning Systems, Weimar, Germany


Principal Investigator:

Prof. Dr.-Ing. Dirk Donath
Chair of Computer Science in Architecture
Bauhaus-University Weimar
Germany


Acknowledgements


Swiss National Science Foundation via project number 100013L_149552 titled "ESUM-analyzing trade-offs between the energy and social performance of urban morphologies."



German Research Foundation (DFG) Research Grant.