Automatic
Commande robuste multicritère - une approche par la théorie des jeux
Publié le
This thesis provides a study of the robustness of controls issued from Game Theory for multiple input and multiple objective systems. Usually strategies with open loop information structure present weak properties of robustness and can be obtained only a posteriori. Some limits are pointed out in examples. Strategies with closed loop information structure are most difficult to design. For Nash strategies, a new algorithm based on the Jack of trust is provided to solve coupled algebraic Riccati type equations. For Stackelberg strategies with closed loop information structure, necessary and sufficient conditions are shown and applied to H2/Hinf mixed control design. At last, an extension to systems including parameter incertainties leads to iterative solving of non convex optimisation problems.