Current research projects
Click here for publications related to this topic.
Large Scale Sensor Networks
The whole concept of sensor networks differs from the classical remote sensing setup because it replaces a single powerful array, sensing a remote area, with a multitude of small individual sensors equipped with a radio interface distributed in the monitored area. Since the communication range is short, sensor systems do not face extreme measurement noise. The challenge faced instead is that of a communication wall.
In addition to the physical data acquisition, a sensor network essentially has to perform four tasks: the data representation (source coding problem), the data communication (channel coding problem), the data processing (signal processing problem) and in most cases respond to the data with feedback control (control problem).
There are two alternative ways to structure the information exchange: in one the sensor information is passed around throughout the network; in the other one the information converges towards a central fusion center. The two scenarios above do not represent the entire network but rather a cluster of nodes within a hierarchical structure.
Data retrieval problems
A basic feature of problems such as that of sensor data acquisition is the inherent redundancy of the data. After all, sensors measure data in a common environment and are there to capture rare and unpredictable changes. One point of view is that of making use of side information which is acquired as the data are pooled together, generalizing the concepts introduced by Slepian and Wolf and Wyner and Ziv. Other approaches try to model the data acquisition as an interpolation problem.
An important aspect to consider is that the separation theorem proved by Shannon for a single communication channel does not hold for the access of correlated sources. Redundancy of the data can be also removed as the information is shuttled throughout the network. These aspects render the problem of data retrieval in sensor networks truly challenging and fascinating.
Our research focuses on finding scalable ways of aggregating data. We study this problem from several perspectives, as illustrated by our work so far.
- Sub-categories in this topic are -
- Distributed source coding
- Data driven scheduling
- Distributed consensus
- Data aggregation in specific applications
[ Back to top ]