This paper presents a positioning system for "people-centric" navigation, which estimates relative positions of surrounding people to help users to find a target person in a crowd of neighbors. Our system, called PCN, employs pedestrian dead reckoning (PDR) and proximity sensing with Bluetooth only using off-the-shelf mobile phones. Utilizing the feature of "group activity" where people naturally form groups moving similarly and together in exhibitions, parties and so on, PCN corrects deviation of distance and direction in PDR. The group information is also helpful to identify the surrounding people in the navigation. A field experiment in a real exhibition with 20 examinees carrying Google Android phones was conducted to show its effectiveness.
The behavior observations on young children in new, first-in-the-life environments have significant implications. We can often uniquely observe a child's unforeseen interaction with the environment and peer-children. It would be not only a piece of discovery but a beginning of an open quest worth exploring. Out-of-classroom activities like kindergarten's field trips are perfect opportunities, but those are quite different from regular classroom activities where the teachers' conventional observation methods are hardly practical. This paper proposes a novel approach to extend the teachers' awareness on the children's field trip behaviors by means of mobile and sensor technology. We adopt the notion of behavioral distribution and singularities. We estimate the children's representative behavioral state in a given context, and study the effect of focusing on the behaviors which are unlikely in this context. We discuss our 14-month collaborative study and various qualitative benefits through multiple deployments on actual kindergarten field trips.
Understanding the effect of pervasive services on user context is critical to many context-aware applications. Detailed descriptions of context-altering services are necessary, and manually adapting them to the local environment is a tedious and error-prone process. We present a method for automatically providing service descriptions by observing and learning from the behavior of a service with respect to its environment. By applying machine learning techniques on the observed behavior, our algorithms produce high quality localized service descriptions. In a series of experiments we show that our approach, which can be easily plugged into existing architectures, facilitates context-awareness without the need for manually added service descriptions.
Commercial home automation systems are becoming increasingly common, affording the opportunity to study technology-augmented homes in real world contexts. In order to understand how these technologies are being integrated into homes and their effects on inhabitants, we conducted a qualitative study involving smart home professionals who provide such technology, people currently in the process of planning or building smart homes, and people currently living in smart homes. We identified motivations for bringing smart technology into homes, and the phases involved in making a home smart. We also explored the varied roles of the smart home inhabitants that emerged during these phases, and several of the challenges and benefits that arise while living in a smart home. Based on these findings we propose open areas and new directions for smart home research.
Emerging electric-drive vehicles, such as hybrid electric vehicles (HEVs) and plug-in HEVs (PHEVs), hold the potential for substantial reduction of fuel consumption and greenhouse gas emissions. User driving behavior, which varies from person to person, can significantly affect (P)HEV operation and the corresponding energy and environmental impacts. Although some studies exist that investigate vehicle performance under different driving behaviors, either directed by vehicle manufacturers or via on-board diagnostic (OBD) devices, they are typically vehicle-specific and require extra device/effort. Moreover, there is no or very limited feedback to an individual driver regarding how his/her personalized driving behavior affects (P)HEV performance. This paper presents a personalized driving behavior monitoring and analysis system for emerging hybrid vehicles. Our design is fully automated and non-intrusive. We propose phone-based multi-modality sensing that captures precise driver–vehicle information through de-noise, calibration, synchronization, and disorientation compensation. We also provide quantitative driver-specific (P)HEV analysis through operation mode classification, energy use and fuel use modeling. The proposed system has been deployed and evaluated with real-world user studies. System evaluation demonstrates highly-accurate (0.88-0.996 correlation and low error) driving behavior sensing, mode classification, energy use and fuel use modeling.
