A Multi-Model Gauss Pseudospectral Optimization Method for Aircraft Trajectories

摘要

An algorithm for solving aircraft trajectory optimization problems using scalable high fidelity simulation models and pseudospectral discretization methods is presented. The resulting parameter optimization problem is solved with an SQP algorithm. One of the problems that arise when using complex simulation models in combination with gradient based optimization algorithms is the generation of an appropriate initial guess. To address this issue two different simulation models are implemented: a mere point-mass simulation model with three degrees of freedom (DoF) and a full nonlinear six degree of freedom dynamic model of the same aircraft. In contrast to the 6-DoF model, the generation of an initial guess for the point-mass model can be performed by a geometric approach that can easily be handled by the user. The optimal solution of the point-mass model will be used as an initial guess for the high fidelity optimization problem. To enable this, an iterated extended Kalman Filter is implemented to adapt the optimal solution of the point-mass model such that it can be used as an initial guess for the 6-DoF optimization. This way, the convergence rate of the overall trajectory optimization problem can be significantly improved. The benefits of the method are illustrated by an air race example involving two dynamic models of an aerobatic aircraft.