Enhanced Fixed Wing UAV Navigation in Extended GNSS Outages using a Vehicle Dynamics Model and Raw GNSS Observables

Abstract

This paper presents the design and testing of an innovative approach to the navigation of fixed wing Unmanned Aerial Vehicles (UAVs) using a Vehicle Dynamic Model (VDM) based, tightly coupled INS/GNSS architecture. The investigation focuses on lowmass (sub 30 kg) and low cost (sub $12,000) applications. Using control inputs from the autopilot combined with knowledge of the vehicle dynamics, an error state extended kalman filter is used to fuse raw GNSS observables and Micro-Electro-Mechanical Systems (MEMS) grade Inertial Measurement Unit (IMU) measurements to provide improved navigation performance in periods that would otherwise be classified as outages in a loosely coupled sense. IMU biases, wind velocity errors and VDM parameter errors are estimated in the filter alongside timing, position, velocity and orientation parameters. Monte Carlo simulations reveal a modest degrading performance with decreased number of satellites in view when using the VDM as opposed to rapid drift in a conventional INS/GNSS approach. A comparison to a typical VDM-based approach operating in a loosely coupled architecture is also made. The proposed VDM/INS/GNSS tightly coupled architecture has been found to offer improved navigation performance with more than 45% improvement in position error with 3 satellites in view as opposed to a loosely coupled VDM/INS/GNSS architecture during a GNSS outage. The estimation error in yaw is improved by 16% with the proposed architecture with 3 satellites as opposed to a VDM scheme operating in a loosely coupled architecture for an outage period lasting 140 seconds. The proposed model-based integration approach using raw GNSS observables is seen to offer improved navigation performance with minimal cost and weight penalty, this approach seems to be a promising method for low cost applications.