Game-theoretic consideration on efficient approximation of maximal output admissible sets by multiple ellipsoids

This paper considers a game-theoretic algorithm that efficiently aligns multiple ellipsoids, the sum of which approximates maximal output admissible sets parameterized in terms of a reference vector. The efficient alignment of the ellipsoids, which is thought of as an optimization maximizing an approximation accuracy, is difficult to be solved in a one-shot manner. Then, the algorithm employs a game and learning achievable to a Nash equilibrium, which helps seek for the better approximation. Finally, this paper presents a numerical example and its computed result in order to make sure that the approximation accuracy is improved, compared to the result by our previous work.