We show that the mixed H2/H? control problem can be efficiently solved using randomized algorithms. Q-parameterization provides a mechanism to search over all stabilizing controllers, and thus gives us the ability to search for H2 minimizing controllers, while still providing stability robustness. Finally, we are able to show that we can get results comparable to a more traditional approach such as gradient search, but in addition, we can solve more complex problems. With very little modification, we are able to deal with multiple objectives, plant uncertainty, and fixed order controllers.
Proceedings of the 2000 American Control Conference
H2 minimizing controllers, Q-parameterization, fixed order controllers
Abdallah, Chaouki T.; Kenneth R. Horspool; R. Scott Erwin; and Peter Dorato. "A Randomized approach to the H2/H? Control Problem via Q-parameterization." Proceedings of the 2000 American Control Conference (2006): 1400-1404. doi:10.1109/ACC.2000.876731.