Trajectory planning in time-varying adverseweather for fixed-wing aircraft using robust model predictive control

Abstract

The avoidance of adverse weather is an inevitable safety-relevant task in aviation. Automated avoidance can help to improve safety and reduce costs in manned and unmanned aviation. For this purpose, a straightforward trajectory planner for a single-source-single-target problem amidst moving obstacles is presented. The functional principle is explained and tested in several scenarios with time-varying polygonal obstacles based on thunderstorm nowcast. It is furthermore applicable to all kinds of nonholonomic planning problems amidst nonlinear moving obstacles, whose motion cannot be described analytically. The presented resolution-complete combinatorial planner uses deterministic state sampling to continuously provide globally near-time-optimal trajectories for the expected case. Inherent uncertainty in the prediction of dynamic environments is implicitly taken into account by a closed feedback loop of a model predictive controller and explicitly by bounded margins. Obstacles are anticipatory avoided while flying inside a mission area. The computed trajectories are time-monotone and meet the nonholonomic turning-flight constraint of fixed-wing aircraft and therefore do not require postprocessing. Furthermore, the planner is capable of considering a time-varying goal and automatically plan holding patterns.