Model-based path prediction for fixed-wing unmanned aircraft using pose estimates

Abstract With the rapid proliferation of small unmanned aircraft systems (sUAS), there is an increasing need for these aircraft to detect and predict each other's motion in order to avoid collisions. This concern arises in addition to the well-established need to detect and avoid manned aircraft. The two threats pose distinct challenges. For example, while a manned aircraft typically travels quite fast compared with a sUAS, its path can be accurately predicted over moderate time intervals using only position measurements and a kinematic particle model. Because sUAS are more maneuverable, and detection horizons can be much shorter, there is a need for more sophisticated prediction methods. One way to improve accuracy is to base predictions on the complete pose (position and attitude) and a higher fidelity model of the threat aircraft's dynamics. As an initial demonstration, we propose an algorithm to predict the path of a small, fixed-wing unmanned aircraft using estimates of this threat aircraft's pose, as might be obtained using visual sensors. To assess the algorithm's performance, predictions using the proposed algorithm are compared with predictions based solely on position data for a large experimental data set. The results indicate that the proposed algorithm outperforms the position-only prediction method.