| Distributed detection has been studied for several decades. Particularly, the design of optimal and suboptimal local decision and fusion rules has been investigated extensively. An issue
that has not been treated in detail in this area is the effects of interference in the communications channel between local decision
makers and the fusion center. Due to bandwidth limitations and the need for flexible transport, sensors in modern sensor networks will often
communicate to the fusion center over shared-access channels. Thus, the effects of interference among sensors is of interest in this context. This issue presents new challenges in the problem of distributed detection, and this paper addresses this challenge.
In particular, in this paper, a distributed detection system infrastructure is provided, and a general multiple access channel model is included to account for imperfect communication between the sensors and the fusion center. The Kullback-Leibler (KL) distance between the
likelihoods of the detection statistics under different hypotheses is used as a performance criterion instead of the probability of error because it can provide more tractable analysis. The performance loss (in terms of KL distance) introduced by imperfect communication over the multiple access channel is analyzed. How to minimize the performance loss (maximize the KL distance) under a total communication power constraint of the sensors is studied, and the optimal power allocation is provided. An interesting finding is that, under certain conditions, a sensor with a worse communication channel is assigned more power than a sensor with a better channel. This is different from the power allocation for wireless communications without detection purpose, such as the
''water-filling'' power allocation scheme, where more power is assigned to better channels.
|
| Xin Zhang received the B.S. and M.S. degree in electrical engineering from Fudan University, Shanghai, China, in 1997 and 2000, respectively, and the Ph.D. degree in electrical engineering from the University of Connecticut, Storrs, CT, in 2005. His research interests include statistical signal processing, detection, and target tracking. He is currently a postdoctoral research fellow at Princeton University, Princeton, NJ. |