| Summary: | 碩士 === 國立交通大學 === 電信工程系所 === 96 === Wireless sensor networks are usually deployed in a harsh
environment. Hence, the performance of the final decision at the
fusion center is not stable because some of the sensors may be
unexpectedly faulty. A straightforward technique to exclude
untrustworth observations from faulty sensors at the fusion center
is to detect them based on the pre-sent training sequences. However,
the transmission of training sequences consumes additional energy,
and shortens the life cycle of the battery-supported sensors.
In this thesis,
we proposed to combine channel estimation and sensor fault
protection in wireless sensor networks in terms of the coding technique.
Simulation annealing is employed to search the best non-linear code
that minimizes the upper error bound.
Quantization is later added in order to eliminate the sudden performance
degradation due to deep fading.
Our results show that
the proposed combined scheme can compete with the
the conventional training-sequence-based fusion in performance
but without the effort of faulty-sensor detection, channel estimation and equalization.
|
|---|
