A numerical efficient algorithm for computing necessary conditions for performance specifications is shown to work on aerospace type problems and provide complementary information to the one provided by Monte Carlo analysis methods. This algorithm addresses the problem of robust trajectory tracking, i.e., determining how far from the nominal the actual trajectory is, under worst-case disturbances and uncertainty. Two norms as measure for the noise signals are used: the undermodeled components gain, and the performance objectives.
The authors develop an algorithm to compute a lower bound on the maximum distance between nominal and actual trajectories. It can accommodate both external disturbances and unmodeled dynamical components in the system as sources of error. Although the algorithm is not proven to converge in general, a numerical study of its behavior shows that it is well behaved when applied to some practical examples. The algorithm is tested on a flight control example.