Fall semester 2020

With the latest technological advancement, see-through smartglasses devices have the potential to improve the daily life of millions of visually impaired individuals. Smartglasses devices could notably support visually impaired individuals for mobility, social interactions and reading. As a new domain of research, many questions remain open in terms of optimal functionalities, human-computer interaction design and potential of feedback modalities.

References

M. Hu, Y. Chen, G. Zhai, Z. Gao, and L. Fan, “AN OVERVIEW OF ASSISTIVE DEVICES FOR BLIND AND VISUALLY IMPAIRED PEOPLE,” Int. J. Robot. Autom., vol. 34, no. 5, 2019.

S. Azenkot and Y. Zhao, “DESIGNING SMARTGLASSES APPLICATIONS FOR PEOPLE WITH LOW VISION,”, ACM SIGACCESS Accessibility and Computing. Issue 119, Nov. 2017.

Reference persons :

• Dr. Simon Ruffieux
• Chiwoong Hwang

During the transition seasons, the decay of natural light during sunset period variates from day to day and can reach insufficient level at different time. As more people are working from home and lightning setup are individual, this could be sometimes unnoticeble when people are concentrating on their work. Sensing technologies which could capture human visual fatigue could assist user to manage their visual comfort. A system that could learn user’s preferences on light control at their domastic workspace and encourage new habits of vision stress management by detecting visual fatigue using eye tracking techniques is an interesting direction to investigate.

References

Christiano, P., Leike, J., Brown, T., Martic, M., Legg, S., Amodei, D. (2017). Deep reinforcement learning from human preferences

De Carli, Michele, Valeria De Giuli, and Roberto Zecchin. Review on visual comfort in office buildings and influence of daylight in productivity. Indoor Air (2008): 17-22.

Borisuit, Apiparn, et al. Effects of realistic office daylighting and electric lighting conditions on visual comfort, alertness and mood. Lighting Research & Technology 47.2 (2015): 192-209.

Reference persons :

• Dr. Hamed S. Alavi
• Sailin Zhong

Data produced automatically by sensors or experimentally in the field of indoor comfort are traditionally aggregated across predefined experimental groups, time and occupants. However, findings produced by this methodology were not found to completely satisfy occupants. One way to overcome this issue would be to extract groups using data-driven approach and to use data as close as raw data as possible without aggregation. The goal of this review is to list the pros and cons of both aggregation and data-driven individualised methods for visual comfort.

References

general review on methods used in indoor comfort https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8812727

example of personal visual comfort https://www.sciencedirect.com/science/article/pii/S0360132314001322

Reference persons :

• Michael Papinutto
• Dr Julien Nembrini

As Air Traffic Flow Management becomes increasingly complex, artificial intelligence (AI)-driven aids will gain in relevance in the future. However, AI-driven aids only prove useful if they are understandable by human users. In a context where Machine Learning (ML) models become more complex and less interpretable than ever, new eXplainable AI(XAI) methods are needed in order to allow the operator to track the behavior of the automation. Nowadays, agents often experience difficulty understanding the behaviour of AI-driven aids, and the benefits offered by automation are balanced by a poor User Experience (UX). This literature review will focus on new methods aimed at improving the cooperartion between Air Traffic Controller and AI-driven systems.

References

Kistan, T.; Gardi, A.; Sabatini, R. Machine Learning and Cognitive Ergonomics in Air Traffic Management: Recent Developments and Considerations for Certification. Aerospace 2018, 5, 103. https://www.mdpi.com/2226-4310/5/4/103#cite

Reference persons :

• Prof. Denis Lalanne
• Raphaël Tuor

Contemporary buildings are fitted with sensors (temperature, presence, light, etc) and actuators (automatic blinds and lights, heating and cooling, etc) which allow dynamic building control to achieve comfort. Such systems generate time series data which can be analyzed to optimize control and/or understand problematic situations. Automatic time series analysis bears the potential to automatically identify non-standard behaviour for notification purpose.

References

Miller, C., Nagy, Z., & Schlueter, A. (2017). A review of unsupervised statistical learning and visual analytics techniques applied to performance analysis of non-residential buildings. Renewable and Sustainable Energy Reviews. https://doi.org/10.1016/j.rser.2017.05.124

Reference person :

• Dr Julien Nembrini