Control architecture is an important problem for the distributed intelligent system. We describe on the evaluation of grouping with plural master cells for a cellular robotic system (CEBOT) as one of the group robotic systems. For the evaluation of the grouping, we use network energy as the estimation function, which we have proposed in a previous paper. By describing the grouping of the network with plural master cells as cores in a subnetwork, we represent the effect for the group structure of the organization or the society. Then for the estimation of the grouping of the network, which is organized by autonomous robotic units called "cells, " we propose the cross-dependence matrix in the cells. The matrix is led from the "task matrix, " which represents the relations between the cells and executing task steps. Finally, we show the simulation results for the example of the organized network by the cellular robotic system. This paper also describes mathematically the efficiency of grouping of the orgnization or the society.