Decentralized ordinal optimization (DOO) for networked systems

This paper proposes a method to implement ordinal optimization in a decentralized way to deal with optimization problems in networked systems. The key idea of DOO is that every sub-system takes samples independently and then cooperates among neighboring sub-systems to find good enough solution. Adaptive learning is also incorporated in DOO to narrow the search space. Numerical examples illustrate the effectiveness and efficiency of DOO.