Data Collection from Resource-Limited Wireless Sensors for Cloud-Based Applications

We consider a wireless sensor network (WSN) where each sensor node samples a random signal, places the digitized data in a buffer, and transmits the data to an access point (AP) through a wireless packet erasure communication link. The AP communicates the data to a processing center (PC) located in the cloud. Our objective is to maximize the delivery of raw data to the cloud for post- processing given the buffer space and communication bandwidth limitations at each sensor. For this system model, we propose an algorithm that is executed at the sensor and summarizes the miminum subset of the incoming data based on the available buffer space and rate of the end-to-end communication link. The PC located in the cloud uses a sequential minimum mean square error (MMSE) estimation algorithm that fuses the summarized and raw data to estimate the random signal. Our algorithm offers a building block for sensing applications that desire the collection of data in raw form.