The performance of target identification can be improved by fusing the data from multiple sensors. Even though distributed fusion has advantages of lower communication bandwidth, less processing at a central location, and increased robustness over centralized fusion, it has to address technical issues such as the conditional dependence of information to be fused by a fusion agent. This paper presents distributed fusion and communication management algorithms for target identification. Information graphs are used to select fusion architectures that minimize the effect of information double counting due to communication. Bayesian networks are used to model the target identification problem and identify the sufficient information that needs to be communicated between processing agents for optimal fusion. Communication strategies are developed to determine when a fusion agent should communicate with another fusion agent. Simulation examples demonstrate the performance of distributed fusion and communication management.
[1]
Chee-Yee Chong.
Graphical Models for Nonlinear Distributed Estimation
,
2004
.
[2]
Stelios C. A. Thomopoulos,et al.
Distributed Fusion Architectures and Algorithms for Target Tracking
,
1997,
Proc. IEEE.
[3]
R. A. Leibler,et al.
On Information and Sufficiency
,
1951
.
[4]
Judea Pearl,et al.
Probabilistic reasoning in intelligent systems - networks of plausible inference
,
1991,
Morgan Kaufmann series in representation and reasoning.
[5]
Finn V. Jensen,et al.
Bayesian Networks and Decision Graphs
,
2001,
Statistics for Engineering and Information Science.
[6]
Hugh Durrant-Whyte,et al.
Communication In General Decentralised Filters And The Coordinated Search Strategy
,
2004
.
[7]
Ken Isaacson,et al.
I I
,
1982
.
[8]
Oliver E. Drummond,et al.
Hybrid sensor fusion algorithm architecture and tracklets
,
1997,
Optics & Photonics.