Hierarchical optimal control with information aggregation for groups including different numbers of agents
Kento Fujita, Daisuke Tsubakino · 2022 American Control Conference (ACC) · 2022
A new quantitative approach to controller design is presented for hierarchical network systems. The distinct feature of the proposed approach is to realize both information aggregation and optimality. In this paper, we consider an optimal control problem for multi-agent systems. Agents, which have the same dynamics, are divided into two groups. The groups include a different number of agents. The control performance is evaluated hierarchically, such as the performance of individual, intra-group and inter-group behaviors. For such systems, we show sufficient conditions under which scalar representative values of each group are exchanged between the groups in the optimal closed-loop system. The conditions are derived based on properties of the matrix algebra. Our results reveal that if the representative value of group and the intra-group performance output are equal, or surprisingly, orthogonal, then an optimal control law inherits the desired hierarchical structure.