Designing Adaptive Intelligent Spaces with Machine Learning as an Interface

by Dr. Asma Naz

Dr. Asma Naz is an Assistant Professor at the Department of Architecture, Bangladesh University of Engineering & Technology (BUET). She has a Bachelor of Architecture (BUET), an M.S. in Visualization Sciences (Texas A&M University, USA) and a Ph.D. in Arts & Technology from UTD (the University of Texas at Dallas, USA). She taught as a senior lecturer at the School of Arts, Technology, and Emerging Communication, UTD. As a designer/Visualization specialist at HKS Inc. (USA), she was involved in several major sports projects. Dr. Naz continues her research in anticipation-guided new forms of intelligent interactive architecture. She had collaborated with Dr. Mihai Nadin (antÉ – Institute for Research in Anticipatory Systems, USA) and Dr. Rainer Malaka (TZI, University of Bremen, Germany), two of the leading researchers working in the field of Interaction Design and Artificial Intelligence respectively. A part of her research on Machine Learning was funded by Hanse-Wissenschaftskolleg/ Institute for Advanced Study, Germany.

“The questions are: How can architectural spaces adapt to assist in enhancing the lives of the people who are forced to stay at and work from home? Would a variable, adaptive environment be helpful to those who interact with it?”

Human beings are adaptive to their changing environments (e.g., seasons, people they interact with, etc.). However, the living conditions of people during the times of the Covid-19 pandemic (in isolation, fear, anxiety, and stress) have given rise to new and exacerbated existing mental and physical health concerns. Many researchers urge to rethink the spaces we live in: not only homes, but also offices, dorms, schools, and healthcare. The questions are: How can architectural spaces adapt to assist in enhancing the lives of the people who are forced to stay at and work from home? Would a variable, adaptive environment be helpful to those who interact with it?

The space generates real-time sensory-perceptive, variable “affective environments” (cozy, calm, exciting, spacious, intimate, etc.) by manipulating space perceptual parameters to accommodate a user’s wants, needs, and desires.

These questions create the premise for the book chapter “Design Driven by Sensory Perceptive Variabilitywritten by Dr. Asma Naz, published in “Human Systems Engineering and Design (IHSED2021): Future Trends and Applications(ISBN: 978-1-7923-8987-0). The chapter presents Dr. Naz’s ongoing research on Intelligent Spaces and Machine Learning (ML). A design concept for an architectural “intelligent” system of adaptive living space is presented where both interacting elements (the human being and the spatial environment) have the capacity to adapt to one another. The space generates real-time sensory-perceptive, variable “affective environments” (cozy, calm, exciting, spacious, intimate, etc.) by manipulating space perceptual parameters to accommodate a user’s wants, needs, and desires. The physical size and shape of the living space remain unchanged.

As space perception is necessarily subjective and relative to context, one of the biggest challenges of this research was to train ML to “understand” the user concepts of “affective environments” and how to generate them as and when needed.

Machine Learning (ML) provides the data that drives perceptual variability. It is the interface for the source of data and enables the living space “learn” to adapt. Data includes aesthetic, demographic, socio-cultural, environmental, as well as a user’s “anticipatory” data accommodating the probabilistic and unpredictable aspects of the living. As space perception is necessarily subjective and relative to context, one of the biggest challenges of this research was to train ML to “understand” the user concepts of “affective environments” and how to generate them as and when needed.

Figure 1. Affective space creation through manipulating properties of light

In the times of Industry 4.0, a paradigm shift in architectural design is inevitable as it collaborates with both Interaction Design and Machine Intelligence.

The design concept presented can also be extended to healthcare—postoperative recovery rooms, nursing care units for persons suffering from physical disabilities, dementia, and other mental impairments—where sensory stimulation is key to treatment. In the times of Industry 4.0, a paradigm shift in architectural design is inevitable as it collaborates with both Interaction Design and Machine Intelligence.

Figure 2. Images of the same scene intended to evoke different emotional impacts: “scariness”
(left) and “coziness” (right).

The design parameters were properties of light: Color, Brightness, and Visible Area. Some data of aesthetic significance were gathered for initiating the design. As mentioned before, training ML is a work in progress. The initial training of ML looked promising, however, a lot more data need to be gathered for accuracy. Data should be collected from people of similar socio-cultural and demographic backgrounds.

Figure 3. Conceptual diagram of Machine Learning as an interface

At a time of the global pandemic, sensory pleasing environment, personalization, and circadian rhythm-based lighting are crucial for emotional and physical well-being.

At a time of the global pandemic, sensory pleasing environment, personalization, and circadian rhythm-based lighting are crucial for emotional and physical well-being. In addition to the emotional effects of the pandemic (fear, anxiety, stress), the system can assist in setting up incentives to facilitate a regular, healthy, active life on a day-to-day basis. This is also true for people who assist patients for long hours (e.g., in ICUs, recuperating rooms, cabins, or care units for patients).

Original Sources: Human Systems Engineering and Design (IHSED 2021)

https://openaccess.cms-conferences.org/#/publications/book/978-1-7923-8987-0

Book Link: http://doi.org/10.54941/ahfe1001094

Chapter Link: http://doi.org/10.54941/ahfe1001161

Author Contact: https://arch.buet.ac.bd/people_faculty/asma-naz/

All contents and images are to the credit of the author/authors