Mixture Generalized Minimum Error Entropy-Based Distributed Lattice Kalman Filter

The Gaussian kernel function-based Minimum Error Entropy (MEE) criterion is effective for special types non-Gaussian noise. However, non-Gaussian noise distributions and shapes are diverse in practice, the traditional MEE methods are difficult to fit non-Gaussian effectively due to the shape parameters of MEE cannot be adjusted. In this paper, the Mixture Generalized Minimum Error Entropy (MGMEE) criterion is proposed by a mixture generalized Gaussian kernel function. Then, a new Mixture Generalized Minimum Error Entropy-based Distributed Lattice Kalman Filter (MGMEE-DLKF) is proposed for multi-sensor nonlinear systems with non-Gaussian noise. The complexity analysis and convergence condition of proposed MGMEE-DLKF algorithm are derived. In the end, the target tracking simulations are verified for systems with mixture Gaussian noise, Rayleigh distribution noise and α−stable distribution noise. The simulation results demonstrate that the proposed filter has the smallest root mean square error.