The use of high sensitivity GPS for initialisation of a foot mounted inertial navigation system

Abstract

This work aims to optimise the use of GNSS for intialising a foot mounted pedestrian INS. The INS error model used in many navigation filters requires the input of an accurate, precise initial position and heading for stable filter performance. However such information is often not available at the start of a navigation session. Often a rough initialisation is performed and position updates are used to refine the INS state vector error estimates over time. This approach is often ineffective in the early portion of a navigation session and can lead to poor filter performance. In this work two innovative methods for using High Sensitivity GPS (HSGPS) observations are developed. In the first, position estimates from the receiver are used to determine rotation and translation parameters for an INS trajectory. This allows the INS to be initialised with an approximate initial position and arbitrary but precise heading. The INS then uses established methods for a foot mounted INS to minimise the relative position drift for a short period of navigation. When sufficient HSGPS positions have been obtained a weighted least squares approach is used to determine the parameters required to rotate and translate the INS trajectory. Once this has been successfully achieved, the INS may change to a traditional loosely coupled GNSS/INS format for longer term navigation. The second method investigates the use of GNSS to provide 'course over ground' measurements to the INS. These estimates can be used by the filter to directly address the problem of heading drift and can therefore be more effective than position updates in a foot mounted pedestrian INS. In this work we test the ability of course over ground measurements to control drift over time. However it is suggested that they can be directly used to determine rotation parameters for initialisation in the style proposed in the first method. To obtain course over ground observations we use the GNSS Kinematic Time Differencing method to difference GPS code pseudorange and carrier phase observations between measurement epochs. We develop an innovative approach to strengthen these estimates by utilising distance traveled information from the INS. It is suggested that in future, the INS information may be used to reduce the number of satellites required to obtain a course over ground estimate. Testing reveals that the course over ground measurement is an effective aiding measurement for a foot mounted INS. It is shown to be superior to HSGPS position aiding when sparse but precise course over ground estimates are available.