An approach for solving aggressive user-specified trajectories in high-fidelity simulations is presented. The objective is to solve for control input histories that satisfy user-defined vehicle states at specified way points in time. In this framework, the user can specify the full rigid-body state of the vehicle, which includes position, velocity, attitude angles, and angular rates. This allows a simulation-user, without expert piloting skills, to push a vehicle model to the limits of its flight envelope, yet in an intuitive and repeatable fashion. These objectives can be accomplished in a variety of ways, including direct and indirect methods for trajectory optimization. This paper presents a direct method employing a gradient-based optimization algorithm within a "direct multiple shooting" formulation. A highly unstable ducted fan rotorcraft aerodynamic model in a 6-degree-of-freedom simulation is the basis for optimizations. Comparisons between optimization of the open-loop aircraft and a rate-stabilized aircraft model are given, showing that unstable models may experience problems with convergence. Using a rate-stabilizing controller in conjunction with the high-fidelity rotorcraft model greatly improved the optimization convergence success rate. The method is demonstrated for several maneuvers of varying complexity.
展开▼
