Joint-Sparse Decentralized Heterogeneous Data Fusion for Target Estimation

In our recent work, we developed a new joint-sparse data-level fusion (JSDLF) approach to integrate heterogeneous sensor data for target discovery. In the approach, the target state space is discretized and data fusion is formulated as a joint sparse signal reconstruction problem, which is solved by using simultaneous orthogonal matching pursuit (SOMP). In our previous work, the joint sparse signal recovery approach has been implemented in a centralized manner. Namely, all the raw sensor data are transmitted to a fusion center, where they are fused to detect and estimate the targets. The drawback of the centralized network is its high communication cost and its lack of robustness, since the global information is stored and processed at a single point, the fusion center. In this paper, several decentralized JSDLF approaches have been developed, that provide exactly the same estimation result at each sensor node as the centralized algorithm does. Further, two distributed database query algorithms, Threshold Algorithm (TA) and Three-Phase Uniform Threshold (TPUT) have been combined with the SOMP algorithm to reduce communication costs. Numerical examples are provided to demonstrate that the proposed decentralized JSDLF approaches obtain excellent performance with accurate target position and velocity estimates to support situational awareness, while at the same time achieving dramatic communication savings.