|
| Figure 1 : Hierarchical Network Architecture for Sensor Testbed |
| Project Summary | Publications | Presentations | Home | Feedback |
Principal Investigator: Sujit Dey
Graduate Students:
Shoubhik Mukhopadhyay,
Debashis Panigrahi
Supported by the
Center for Wireless Communications
and UC Discovery Grant
Reliability in Wireless Sensor Networks:
Future wireless sensor networks promise a truly mobile and ubiquitous information technology era, enabling automated data gathering, communication, and processing. However, wireless sensor data can be subject to various sources of unreliability, e.g. soft errors in the sensing and processing circuits, increased failure rates due to often remote deployment without maintenance, tampering with the data, and channel noise and interference during wireless transmission. Since data is collected, transmitted and processed automatically, an obvious concern in large scale deployment of wireless sensor networks in several data-critical segments like medicine, emergency response, and security, is ensuring that data collected and communicated by sensors is authentic and reliable. Existing techniques for reliability and error-correction, like circuit hardening, tamper-resistant encryption, and channel coding incur significant overheads in sensor silicon area, energy consumed, and communication costs. Our goal is to identify low-cost error correction techniques for sensor applications that will be able to provide reliable sensor data aggregation against any of the different sources of errors in a sensor network. Our proposed reliable data aggregation techniques, deployed in sensor micro-gateways, exploit the a priori knowledge of sensor data and application semantics, the vulnerability of the application to various errors, and also the nature of spatial and temporal correlation of sensor data. We are developing a comprehensive sensor network test-bed, which we are using to collect real sensor data, and validate our low-cost, reliable data aggregation techniques. Preliminary studies have shown the ability to provide reliable sensor data aggregation, without any circuit hardening, encryption or channel coding costs in the sensors.
The architecture of the network testbed (Figure 1) which will be used consists of a two level hierarchy of different types of sensors as well as infrastructural nodes called micro-gateways. There will be various types of sensor nodes in the lower level, their complexity only dependent upon the type of sensed data. The next level will consist of the micro-gateways which will be responsible for computation-intensive tasks like data aggregation, reliability functions, and sensor scheduling and configuration.
|
|
| Figure 1 : Hierarchical Network Architecture for Sensor Testbed |
Low Cost Data Aggregation:
For networks which include high volume data sources such as video sensors, the amount of data generated by a large number of such sources will make the bandwidth as well as energy costs prohibitive. However, such data are also charaterized by spatial and temporal redundancies within as well as across nodes. We assume that each sensor node has a different set of multiple configurations it can be operated at. Each of these configurations provides a different level of resolution of sensed data, with corresponding operational costs and data volume. Now, the overall cost of operation as well as the quality of the aggregated observations is dependent on the configurations at which each sensor in the network is being operated over time. Which schedule of these configurations provides an optimal trade-off depends on the spatial and temporal correlation within the sensed data. The goal of our research is to devise distributed global algorithms for calculating such schedules based on current and previous observations. These algorithms will later be implemented within the micro-gateways in the network testbed mentioned above.