A Suboptimal CI Algorithm and its Application to Distributed Target Tracking

When the fused variables are independent or the statistics of variables are known perfectly, Kalman filter is rigorous and yields minimum mean squared error estimate. But in most situations, it is impossible to guarantee that the fused variables are independent and even might be highly correlated. In that case, it is possible to "over estimate" the statistics. CI algorithm can achieve a consistent estimation, when the correlation between the fused variables are unknown. However the caculation of optimal ω of CI costs lots of time. A simple way to calculate a suboptimal ω is proposed to reduce the time complexity and get a sub optimal estimation. According to the geometrical explanation of CI, a simple calculation equation ω of is given. And the functions of three kinds of fusion algorithm are illustrated in an application of decentralized estimation with circle topology, where it is impossible to consistently use a Kalman filter or other algorithms that need independent constraints.