Wireless sensor networks provide an attractive approach to spatially monitoring environments. Wireless technology makes these systems relatively flexible, but also places heavy demands on energy consumption for communications. This raises a fundamental trade-off: using higher densities of sensors provides more measurements, higher resolution and better accuracy, but requires more communications and processing. This paper proposes a new approach, called "back-casting," which can significantly reduce communications and energy consumption while maintaining high accuracy. Back-casting operates by first having a small subset of the wireless sensors communicate their information to a fusion center. This provides an initial estimate of the environment being sensed, and guides the allocation of additional network resources. Specifically, the fusion center backcasts information based on the initial estimate to the network at large, selectively activating additional sensor nodes in order to achieve a target error level. The key idea is that the initial estimate can detect correlations in the environment, indicating that many sensors may not need to be activated by the fusion center. Thus, adaptive sampling can save energy compared to dense, non-adaptive sampling. This method is theoretically analyzed in the context of field estimation and it is shown that the energy savings can be quite significant compared to conventional approaches. For example, when sensing a piecewise smooth field with an array of 100 /spl times/ 100 sensors, adaptive sampling can reduce the energy consumption by roughly a factor of 10 while providing the same accuracy achievable if all sensors were activated.
[1]
Urbashi Mitra,et al.
Boundary Estimation in Sensor Networks: Theory and Methods
,
2003,
IPSN.
[2]
R. Nowak,et al.
Multiscale likelihood analysis and complexity penalized estimation
,
2004,
math/0406424.
[3]
E. Candès,et al.
Curvelets: A Surprisingly Effective Nonadaptive Representation for Objects with Edges
,
2000
.
[4]
D. Donoho.
Wedgelets: nearly minimax estimation of edges
,
1999
.
[5]
A. P. Korostelev,et al.
MiniMax Methods for Image Reconstruction
,
1993
.
[6]
Leo Breiman,et al.
Classification and Regression Trees
,
1984
.
[7]
Urbashi Mitra,et al.
Estimating inhomogeneous fields using wireless sensor networks
,
2004,
IEEE Journal on Selected Areas in Communications.
[8]
A. Tsybakov,et al.
Minimax theory of image reconstruction
,
1993
.
[9]
Robert D. Nowak,et al.
Platelets: a multiscale approach for recovering edges and surfaces in photon-limited medical imaging
,
2003,
IEEE Transactions on Medical Imaging.