Intelligent Avionics with Advanced Clustering

In this paper, we consider tracking targets using multiple distributed sensor platforms. Rather than sending the tracks to a central location, such as a command and control center where information is exchanged between platforms, we consider a distributed solution. While fixed position single sensor tracking of a single target is considered straightforward, multiple sensors on the different platforms with overlapping coverage is complex because duplicate tracking data is generated for the same targets. Redundant information generates network messages that in turn overload the network performance, and may result in traffic congestion on limited avionic bandwidth wireless links that prevents time critical data from reaching its destination. In our approach, we identify similar tracking data to be distributed and send only one copy of the message. In order to identify redundant data, we use a clustering algorithm to evaluate the large volumes of sensor information. Distributed multiple target tracking (MTT) combines track observations from different sensors to identify the same target. Massive computation and communication is required for distributed real time MTT. In this paper, the K-means clustering algorithm is used to aggregate redundant tracks. Software simulations using Matlab and emulation tests using Emulab show significant improvement of the information quality by using clustering. An MTT system was prototyped with FPGA hardware to cluster high volumes of data with low latency in real time at the network layer.