Towards new notions of reliability for smart homes

Reliability in systems traditionally refers to notions of fault-tolerance and security. Yet, for systems that involve significant human interaction – smart homes being an example – the question of whether the system can be relied on in daily activities, needs us to go beyond the traditional definitions of reliability. In this thesis, we argue that human-facing aspects of reliability should be considered as a first-class citizen when designing human-facing systems and reasoning about the reliability of such systems. In particular, we focus on smart home and IoT settings. Our works straddle the worlds of both systems design and human-computer interaction. We apply research methodologies from both research areas and conduct multi-pronged evaluations covering both systems metrics and human-facing aspects. We show how systems can provide responsiveness and user satisfaction to smart home users, by: i) designing new systems for fast and low-cost failure detection in IoT networks, ii) designing new systems that provide novel notions of human-facing reliability in smart homes operating with routines, and iii) human user studies of user preferences for resolving conflicts among smart home routines, which lead to our discovery that human values underlie preferences and our proposal of a tagging solution for users to specify values. Concretely, first we build an efficient fully distributed failure detector Medley, and its successor Medley-F, to provide system-aspect reliability support for IoT applications to be built atop. This is an example of traditional system-facing reliability. We evaluate Medley and Medley-F through mathematical analysis, simulations, and Raspberry Pi deployments. Second, we introduce new human-facing notions of smart home reliability, including atomicity (SafeHome), visibility (SafeHome), and a value-based tagging system. We perform both system evaluations (for SafeHome) and user evaluations (for routine conflicts and the tagging system) to understand new aspects of human-facing reliability. Our user studies highlight that it is important for future smart home system designers to explicitly consider user needs and wants when designing smart home management systems. Systems design (as well as UI) should consciously reason about mental models.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Yang, Rui

Contributors dc:contributor

• Gupta, Indranil
• Karahalios, Karrie
• Vasisht, Deepak
• Shu, Yuanchao

• Reliability
• Smart Home
• Internet of Things (IoT)
• Failure Detector
• User Satisfaction

Statement dc:rights

• Copyright 2022 Rui Yang

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/115561

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F115561