Predictive Data Delivery to Mobile Users Through Mobility Learning in Wireless Sensor Networks

We consider applications, such as indoor navigation, evacuation, or targeted advertising, where mobile users equipped with a smartphone-class device require access to sensor network data measured in their proximity. Specifically, we focus on efficient communication protocols between static sensors and users with changing location. Our main contribution is to predict a set of possible future paths for each user and store data at sensor nodes with which the user is likely to associate. We use historical data of radio connectivity between users and static sensor nodes to predict the future user-node associations and propose a network optimization process, i.e., data stashing, which uses the predictions to minimize network and energy overheads of packet transmissions. We show that data stashing significantly decreases routing cost for delivering data from stationary sensor nodes to multiple mobile users compared with routing protocols where sensor nodes immediately deliver data to the last known association nodes of mobile users. We also show that the scheme provides better load balancing, avoiding collisions and consuming energy resources evenly throughout the network, leading to longer overall network lifetime. Finally, we demonstrate that even limited knowledge of the location of future users can lead to significant improvements in routing performance.