Pipelined Particle Filter with Non-Observability Measure for On-Board Navigation with MAVs

In this paper, a particle filter (PF) is proposed to estimate attitude and velocity on board a mini aerial vehicle (MAV) using only GPS measurements in the re-weighting step. The attitude observability of the system heavily depends on the vehicle trajectory. Thus, a measure of non-observability is introduced to track the uncertainty in the attitude estimate. For the nonlinear estimation problem at hand, the use of a PF instead of a classical extended Kalman filter has shown to be advantageous in terms of robustness. However, PFs are computationally very demanding and thus, are usually not applicable for navigation on board MAVs. To overcome this problem, an on-board computer system (OCS) has been developed which comprises a field programmable gate array (FPGA) as coprocessor. To get a quantitative measure on the performance of the proposed algorithm, the FPGA-based OCS running a pipelined PF for attitude and velocity estimation has been flight-tested on a civil aviation aircraft equipped with a high-precision attitude and heading reference system. Further tests have been carried out with a small fixed-wing MAV, where the pipelined PF has been used as attitude reference in a simple waypoint navigation scenario.