Information fusion using adaptive MCMC method in integrated navigation system

The integrated navigation system is a nonlinear and non-Gaussian system. Based on these features, an Adaptive MCMC (Markov Chain Monte Carlo) Particle Filter was proposed, in which the real time amount of particles was adaptively adjusted online. The MCMC process in the Adaptive MCMC Particle Filter was used to maintain diversity of particles, avoiding sample degeneracy and impoverishment; and the KLD sampling approach was imported in order to adaptively choose the number of particles according to the state uncertainty, improving efficiency. In the last part, an experiment in SINS/GPS Integrated Navigation System was given. The experiment shows a great reduction of calculate amount and time when using Adaptive MCMC Particle Filter instead of standard MCMC Particle Filter with the same accuracy.