Assessment of VDM-based autonomous navigation of a UAV under operational conditions

This paper presents extensions and practical realization of a previously proposed novel approach to navigation and sensor integration for small unmanned aerial vehicles (UAV). The proposed approach employs vehicle dynamic model (VDM) as process model within navigation system, and treats data from other sensors such as inertial measurement unit (IMU), barometric altimeter, and global navigation satellite system (GNSS) receiver as observations within the system. In comparison to conventional approach that employs inertial navigation system (INS) as process model, employing VDM requires no added hardware, yet significantly improves navigation performance, especially in case of GNSS outages. Experimental results from a real flight on a custom made fixed-wing UAV, as well as Monte Carlo simulation results, reveal improvements of 1 to 2 orders of magnitude in navigation accuracy during GNSS outages of 3 minutes' duration. This is a prerequisite for safer navigation without exteroceptive sensors. Uncertainty levels are predicted consistently within the filter, and a discussion on observability based on covariance matrix analysis is presented. Computation time is also compared to conventional INS-based approach.