Density estimation for out-of-range events on personal mobile devices

Abstract

Over the years, personal mobile devices have obtained increasing capabilities to concurrently connect to surrounding networks and nearby other devices. Good knowledge on the dynamics in the availability of these heterogeneous entities constitutes essential input for various data communication problems, ranging from the adaptation of applications on mobile devices to multi-homed situations, to the optimization of routing protocols for delay tolerant networks. In this paper, we focus on a method for the prediction in time of the loss in visibility of currently in-range network entities, as observed on a personal mobile device. We are interested in estimating the full probability density function of the time of these out-of-range events, as this allows us to ask arbitrary questions such as: what is the probability of losing connection X in the next Y minutes? To do so, we model the mobility of the user by applying kernel density estimation on previously observed mobility traces collected during a user experiment we ran with 12 participants in a six week period, logging cellular, 802.11 wireless LAN, Bluetooth, and various other events on the participant's mobile device.

Predicting mobility events on personal devices

“…The CoSphere data set [14] was collected with the middleware software implementing the CoSphere NAL [15], gathering data on cellular, 802.11 wireless LAN, and Bluetooth networks in range. The duration of this trial was approximately one month and had 12 participants carrying their mobile device in their everyday activities.…”

“…The stream data contains interruptions caused by device resets during the collection of the data [14]. We have preprocessed the stream such that short interruptions less than 15 minutes are considered to not change the stream state: we assume no events happen during these short interruptions.…”

Reliability of Bluetooth-based connectivity traces for the characterization of human interaction

“…Recently, other interaction experiments have been deployed in specific environments, such as the one carried out during a roller tour in Paris with rollerbladers wearing Bluetooth motes [17] or the massive Bluetooth data collection in the city of Bath [18]. There are also research projects that have combined multiple radio communication technologies to collect human traces [19,20].…”

Understanding the complexities of Bluetooth for representing real-life social networks

“…Numerous research have used BT data for recognizing social patterns, inferring social relationships and creating networking structures [1][2][3][4][5][6]. In addition, the versatility of BT data have been demonstrated by leveraging BT for context-oriented opportunistic networking applications and epidemic modeling [20][21][22], or even influencing mental health [7].…”