Distributed asynchronous multiple sensor fusion with nonlinear multiple models

This article studies a distributed estimation fusion problem with nonlinear multiple dynamic models under asynchronous multi-rate multi-sensor conditions. Such conditions allow for more comprehensive and various dynamic modes and reflect more practical sensor environments than have previously been studied. Whereas other estimation fusion algorithms are limited in that they are based on an assumption of either a linear dynamic model or synchronous sensors, the present algorithms proposed in this study can handle both nonlinear multiple models and asynchronous sensor observations. A distributed fusion algorithm for a single nonlinear model with asynchronous multiple sensors is proposed using the fusion of the information matrix and information state contribution, which have been reconstructed with statistical linear error propagation based on unscented transformation. The distributed fusion algorithm is then applied to multiple nonlinear models using an interacting multiple model (IMM) approach. In this study, one-step prediction for each dynamic model included in the IMM is performed instead of prediction of the fused estimates. This accounts for the lack of a global model for IMM. Then the information matrix, information state contribution, and mode likelihood function for each model obtained from each local sensor are fused. Simulation studies for tracking with both a single comprehensive nonlinear model and multiple models using asynchronous multiple sensors are conducted to illustrate the proposed algorithms.