Enabling Accurate and Energy-Efficient Context-Aware Systems for Smart Objects using Cellular Signals

The Internet of Things (IoT) paradigm aims to interconnect a variety of heterogeneous Smart Objects (e.g., sensors, smart devices, home automation equipment) using Machine-to-Machine communications. Smart devices have become one of the primary ways for people to access entertainment and other business applications, both inside and outside of their homes. This has led to two significant problems: substantial increase in monthly wireless data usage, and a rapid drain in smart phone battery life. Another recent trend with small form-factors in devices has lead to a bulk of the device components fused together using adhesives without being exposed to outside world (e.g., battery is glued to panel case or screen without exposing the circuit terminals). This prevents researchers from measuring energy consumption ratings for the different sub-systems in the phone using power monitoring devices.;Smart devices that provide health monitoring, smart home and workplace, enterprise device management, and many others need to constantly sense their context and communicate with the network to collaborate with others. Mobile applications that provide location-specific services require either the absolute or logical location of users in indoor settings. Identifying the context of a user (e.g., in front of the store, suits section, billing counter, home, office, conference room) in a timely and energy-efficient manner is important for the applications to disburse appropriate deals or activate a set of device-specific policies. In all these cases, though sub-meter level accuracy is not required or expected, a practical and an infrastructure-independent solution which can be easily deployed in real world is highly preferred.;In this research, we first analyze the detailed statistical properties of cellular signals in indoor environments and construct a reliable database of cellular signal signatures for different indoor locations. We show that it is feasible to accurately distinguish between neighbouring indoor locations in a reliable and energy-efficient manner. We then profile the energy usage of Wi-Fi in mobile devices under different device screen activation scenarios and quantify the energy wastage due to unnecessary scan and association events under poor link conditions, which to the best of our knowledge has not been reported in previous literature. In our first work, iSha, we develop a fine-grained energy consumption analyzer system to estimate the energy consumption values of specific sub-components in smart devices which eliminates the need for specialized hardware power monitoring equipments.;In our second work, PRiSM, we develop a novel and light-weight signature matching system to automatically discover Wi-Fi hotspots without turning on the Wi-Fi interface in the smart device. It uses signal strengths received from cellular base stations to statistically predict the presence of Wi-Fi and connects directly to the hotspot without scanning. The system continuously learns based on user movement behaviours and auto-tunes its parameters accordingly. Hence, PRiSM, provides a practical and infrastructure-independent system to maximize Wi-Fi data offloading and simultaneously minimize Wi-Fi sensing costs.;In our final work, PILS, we develop a indoor localization system which logically maps the contextual information of the smart device with a specific indoor location using cellular multihoming. We utilize a variety of back-channel parameters such as Received Signal Code Power (RSCP) from 3G radio cellular systems, Reference Signal Received Power (RSRP) and Reference Signal Received Quality (RSRQ) from 4G radio cellular systems in addition to Received Signal Strength (RSS) values from 2G radio cellular systems. We show the effects on location accuracy with using only connected base stations and with neighbouring base stations, self-sourced data and crowd-sourced data. We also show that by choosing a combination of signals from different cellular radio technologies specific to different locations provide better location accuracy than relying on one single radio technology for all indoor locations.;In short, we aim to address three important challenges in ubiquitous and pervasive mobile computing: maximal data offloading from cellular networks to Wi-Fi with minimal energy consumption, fine-grained energy consumption analysis for small form-factor devices, and cost-effective and infrastructure-independent indoor localization system for wide-area IoT networks. We show the effectiveness of our solutions with working system prototypes and real world data analysis results. We also show that our solution methodologies are robust and applicable to all major mobile computing platforms.