Some Issues On Distributed Detection In Wireless Sensor Networks

In the past ten years, wireless sensor networks have attracted many interests in the academic and industry areas. A typical wireless sensor network is consisted of a special phenomenon which is under monitor, some randomly scattered local sensor nodes, and a powerful fusion center. Moreover, the local sensor node just has a limited amount of power and computing resource, so most of the computing and fusing task is undertook by the fusion center. For a distributed detection problem, first many local sensor nodes detect and sense the special phenomena or event, then they transmit their own detecting result to the fusion center through a wireless channel, and the fusion center synthesis all the received results from the local sensor nodes, which are affected by the channel fading and noise, to make the final decision. Since one of the most important functions of the wireless sensor network is to detect the special event and some special parameter, so how to design a efficient distributed detection system is a core topic. There are two main factor which affect the performance of the distributed detection algorithm: one is the local sensing performance of the local sensor nodes which is determined by the noise and the interference of the monitored region, the other is the channel quality through which the local detected results are transmitted to the fusion center and the fusion algorithms carried by the fusion center to make the final decision. Among the two factors, the fading channel's property is the main feature making the distributed detection problem in wireless sensor network distinct with the traditional data fusion problem. All the research on the this topic is carried out to exploit the two factors, especially the fading properties of the wireless channel, which is the hot topic in the academic research in this topic, we introduce some basic background of the problem, also include some basic detection theory,multiple input multiple output communication theory,orthogonal space time coding theory and virtual MIMO concept and then we focus the specific distributed detection algorithms and get some novel results as follows:(1) A novel robust distributed detection algorithm is proposed, which does not need any knowledge of the channel amplitude. Specially, we got a theorem to summarize the detector's structure. This is a model-free detector, that is, no channel knowledge is needed. In this way, we avoid the flaws of the current detectors, which need some channel knowledge to aid the detection task, so our algorithm is highly roust to all kinds of channel, which is very important to some special case where the channel is highly complex and bad and no channel information is available to the fusion center.(2) A novel orthogonal space time code aided distributed detection algorithm is proposed. We introduce the space dimension to combat the fading channel. Since all the current detectors are working is a no-diversity based channel, so this is a bottleneck for the whole distributed detector system. Thanks to the special orthogonal structure introduced by the space time code scheme, we can convert the new problem to a traditional problem and get the corresponding solution. But the distinct difference is our new channel model exhibits a higher diversity order than the counterpart of the traditional case. So we can reconstruct the fading channel and improve the whole detection performance.