Rao-Blackwellised Inertial-SLAM with Partitioned Vehicle Subspace

Abstract

This paper presents methods which enable the Rao-Blackwellised (R-B) particle filtering technique to be applicable for the airborne simultaneous localisation and mapping problem. Although R-B filter has been successfully applied to mobile/ground vehicles, its extension to flying vehicles has been impractical due to the high dimensionality involved in inertial navigation system (INS). To overcome this problem, the full INS state is further partitioned into an external state (vehicle pose) and an internal state (navigation and sensor calibration), with a particle filter being applied only to the external state. The computational complexity is further reduced by developing a hybrid R-B Inertial-SLAM. Simulation results will be presented with simulated flight data, showing reliable performances during loop-closures.