Enabling and Advancing Human and Probabilistic Context Awareness for Smart Facilities and Elder Care
pdf
Abstract
Knowledge of the locations of people is critical to many applications, such as smart buildings, in-home elder care, smart authorization, and emergency response. However, existing applications assume that all people are radio- tagged and thus actively participate in the localization effort. In this project, we use radio tomography (RT), i.e. the changes in radio signal strength (RSS) measurements on several static links over time, to estimate where people and moving objects are located. RT presents several advantages: (a) its ability to see through nonmetal walls, darkness, and smoke, (b) the fact that it is a non-visual technology not having the same privacy concerns as video cameras, (c) its ability to track multiple people simultaneously moving in the same area, and (d) its low cost (i.e., ZigBee system-on-chips cost few dollars, compared to very expensive ultrawideband (UWB) radars). By knowing where people are located, including those who do not participate in the localization effort, we will be able to measure human context more completely and effectively.
Context-aware computing systems to date generally assume for the sake of simplicity that the contexts they use are completely accurate, resulting in ineffective and at times faulty systems. In real-world scenarios, context information is often ambiguous and uncertain, yet humans continually resolve and reason despite the uncertainty. Thus, there is a significant added value in representing and displaying uncertainty and confidence information. In this project, we develop new context-aware middleware which is able to: (a) maintain and share uncertainty information with consumers of context, both at the application and middleware levels, (b) represent multiple possible resolutions of the context, (c) when possible remove the uncertainty in context, and (d) perform sensor fusion in the face of uncertain context. These additions represent an important shift in the design of context-aware middleware and expand human context awareness in computing systems.
Award ID: 1035565
Submitted by Neal Patwari
on
Abstract
Knowledge of the locations of people is critical to many applications, such as smart buildings, in-home elder care, smart authorization, and emergency response. However, existing applications assume that all people are radio- tagged and thus actively participate in the localization effort. In this project, we use radio tomography (RT), i.e. the changes in radio signal strength (RSS) measurements on several static links over time, to estimate where people and moving objects are located. RT presents several advantages: (a) its ability to see through nonmetal walls, darkness, and smoke, (b) the fact that it is a non-visual technology not having the same privacy concerns as video cameras, (c) its ability to track multiple people simultaneously moving in the same area, and (d) its low cost (i.e., ZigBee system-on-chips cost few dollars, compared to very expensive ultrawideband (UWB) radars). By knowing where people are located, including those who do not participate in the localization effort, we will be able to measure human context more completely and effectively.
Context-aware computing systems to date generally assume for the sake of simplicity that the contexts they use are completely accurate, resulting in ineffective and at times faulty systems. In real-world scenarios, context information is often ambiguous and uncertain, yet humans continually resolve and reason despite the uncertainty. Thus, there is a significant added value in representing and displaying uncertainty and confidence information. In this project, we develop new context-aware middleware which is able to: (a) maintain and share uncertainty information with consumers of context, both at the application and middleware levels, (b) represent multiple possible resolutions of the context, (c) when possible remove the uncertainty in context, and (d) perform sensor fusion in the face of uncertain context. These additions represent an important shift in the design of context-aware middleware and expand human context awareness in computing systems.
Award ID: 1035565