查看文章

In sensor networks, it is the Quality of Information (QoI) delivered to the end user that is of primary interest. In general, measurements from different sensor nodes do not contribute equally to the QoI because of differing sensing modalities, node locations, noise levels, sensing channel conditions, fault status, and physical process dynamics. In addition, metrics of QoI are highly application dependent, such as probability of detection of an event or fidelity of reconstruction of a spatio-temporal process. Despite these considerations, traditional data dissemination protocols in sensor networks have been designed with a focus on metrics such as throughput, packet delivery ratio, latency, and fair division of bandwidth, and are thus oblivious to the importance and quality of sensor data and the target application. In this paper, we argue for sensor network protocols that are cognizant of and use feedback from the sensor fusion algorithms to explicitly optimize for application-relevant QoI metrics during network resource allocation decisions. Through analysis and simulation we demonstrate the application-level performance benefits accruing from such a QoI-aware approach to network resource management in the context of a centralized sensor rate selection mechanism for an event detection application scenario.