Temperature Map Recovery Based on Compressive Sensing for Large-Scale Wireless Sensor Networks

Large-Scale Wireless Sensor Networks are widely applied into monitoring indoor and outdoor special events, such as spectrum estimation and temperature sensing. The way of sensing is used to deploy plenty of sensor nodes in the field. However, after analyzing the data we collected from WSNs, we´ve found that events happened in the same area has presented clustering stably. Instead of randomly showing up all around, events tend to emerge from some individual points. For that reason, the nodes that distributed indiscriminately certainly have redundancies. And these redundancies will bring unnecessarily energy cost. In consider of that, by combining the compressive sensing and matrix completion theory and analyzing the data, we have found that as long as the data collected from WSNs is close to low-rank matrix criteria, at most 88% of nodes can be cut down and 90% energy will be saved.