Gaussian process-based Bayesian non-linear filtering for online target tracking

This study deals with the problem of online target tracking based on Gaussian process regression and deterministic sampling. A general mathematical formulation of the non-linear system identification problem is constructed based on the Gaussian process model, in which the computational efficiency is improved with a sparsification method, and the regular Gaussian process regression algorithm is extended to the online case where the learned surrogate model is able to be updated recursively. After that, the problem of non-linear transformation with stochastic uncertainty is considered, in which the mean and covariance of a random variable is transformed based on a Gaussian process model. Then the proposed non-linear transformation method is applied to solve the Bayesian non-linear filtering problem with unknown dynamics and measurement models. In the end, extensive Monte-Carlo simulations are conducted to verify the performance of the Gaussian process-based non-linear filtering method in a reentry target tracking scenario.