Real-time implementation of mixture particle filter for 3D RISS/GPS integrated navigation solution

摘要

An optimised real-time realisation of a mixture particle filter (PF) integrated 3D navigation system for land vehicles based on Global Positioning System (GPS) and Reduced Inertial Sensors System (RISS) is introduced. The PF is a nonlinear filter that can handle errors and uncertainties. Although a mixture PF decreases the number of particles compared to a sampling/importance resampling (SIR) PF, the implementation on embedded systems needs further optimisation to run at higher rates. Introduced is an optimised realtime implementation of a mixture PF on a 600 MHz ARM processor. The optimisation is based on fast median-cut clustering to reduce the complexity of search in the weighting step. The proposed real-time system was tested on a real mobile robot trajectory, showing fast and accurate performance.