AFRL engineers develop advanced information management capabilities to monitor wireless sensor networks.

Recent advances in the development of microsensors, microprocessors, information fusion algorithms, and ad hoc networking have led to increasingly capable wireless sensor networks. These networks, when deployed to monitor an urban area, show great promise in enhancing warfighter situational awareness. However, delivering the sensor network’s collected information back to the proper decision makers is one network capability that still requires improvement. To bridge this gap between the tactical operations center and multiple wireless sensor networks distributed across a city, engineers must create a system-of-systems architecture. This architecture must permit a warfighter to receive near-real-time sensor information from an out-of-theater operating post, whether a mile or an ocean away. Research accomplished in efforts such as the Defense Advanced Research Projects Agency’s (DARPA) Information Exploitation Officesponsored Networked Embedded Systems Technology (NEST) program has provided information gathering algorithms for wireless sensor networks that are independent of the hardware platform on which they run. Nevertheless, these networks have no means for publishing the massive amounts of information to the Global Information Grid (GIG). To address this publication requirement, AFRL engineers have begun integrating NEST technologies with the Joint Battlespace Infosphere (JBI).^{1, 2} They recently developed a proof-of-concept demonstration of this idea for Scientific Advisory Board (SAB) review. In this demonstration, they integrated a tracking application developed for the NEST program with the AFRLdeveloped JBI Reference Implementation and showcased the resultant capability to connect low-level information gatherers to high-level information distributors.

To visually convey the benefits of this combined NEST/JBI architecture to the SAB, the AFRL engineers deployed a small wireless sensor network. The NEST application included a display screen showing the overhead layout of this network, and as a network node’s sensors detected objects, the screen displayed the estimated location of each object. To reduce false alarm rates, the engineers divided the network into clusters and calculated a corresponding confidence level for each cluster. The team assigned one node from each cluster as a cluster master; the role of this designated node was to aggregate information from its neighboring nodes and report to the local data collection station the number of nodes in its cluster that sensed the intruding object. If the number of nodes detecting the intruder within a certain time period was above a predefined threshold, the system accepted the information from that cluster as being useful. The figure illustrates the network architecture and information flow established for the SAB review. As depicted, the engineers modified the NEST application (representing an in-field tactical sensor network) to publish network events, such as object detection, to the JBI. They configured another application to (1) duplicate the overhead of the network view, (2) connect to the JBI as a subscriber to the NEST application’s published network events, and (3) display subscribed events to a remote operator monitoring the battlespace. As illustrated, a local data collection station is necessary in the deployment as a means of translating raw sensor messages into meaningful information objects for the JBI. Connectivity between this local station and the JBI server is subject to the constraints of the specific environment unique to each deployment. During the NEST/JBI demonstration, the engineers executed the tracking application and sent—via a wireless connection meeting the 802.11b communications standard—network events from the customized NEST application, through the JBI, and to the network monitoring application.

The number of messages passed from the sensor network to the JBI depends on numerous factors, as exemplified by the following real-world example. In December 2004, DARPA, the NEST program team, and The Ohio State University conducted the Extreme-Scale Wireless Networking (ExScal) demonstration, which consisted of roughly 1,000 sensor nodes in a nonaggregating, multitiered network.³ In the ExScal environment, sensor data is forwarded to a single network access point, which then receives and stores the information on a local computer. For a particularly large event, such as an earthquake, it is possible that every node in the network will send at least a single message. Assuming 100% sensor reliability, the earthquake would generate a throughput of 1,000 information objects (1 per node) in a very short period of time. Such volume easily overloads the computer and can lead to dropped messages.