Vehicle Model Aided Inertial Navigation for a UAV using Low-cost Sensors

This paper considers two different methods of using a dynamic vehicle model in order to aid pose estimates provided by an Inertial Navigation System (INS) for a Un-manned Aerial Vehicle (UAV). We consider low-cost inertial sensors in which errors from sensor noise, bias and scale-factor errors cause a significant growth in pose estimate errors when the navigation system is un-aided by external positioning information such as from GPS or terrain observations. It is shown that an accurate dynamic vehicle model can considerably reduce the error growth in vehicle pose estimates by enhancing the observability of the inertial sensor error sources. Furthermore we analyse the effectiveness of using the vehicle model information given that there are errors in the vehicle model.