A consistent track-to-track fusion method based on copula theory

This paper addresses the problem of distributed fusion when the conditional independence assumptions on sensor measurements or local estimates are notmet. A new data fusion algorithmcalled Copula fusion is presented.The proposedmethod is grounded on Copula statistical modeling and Bayesian analysis.The primary advantage of the Copula-based methodology is that it could reveal the unknown correlation that allows one to build joint probability distributions with potentially arbitrary underlyingmarginals and a desired intermodal dependence.The proposed fusion algorithm requires no a priori knowledge of communications patterns or network connectivity.The simulation results show that the Copula fusion brings a consistent estimate for a wide range of process noises.