Aiming at the poor stability of the final convergence of multi-agent systems in the process of pinning control, we design an influence degree matrix and reconstruct an intermediate centrality algorithm to complete the selection of pinning control nodes. Firstly, an influence degree matrix is calculated according to degree distribution of subnets. Then, the influence degree matrix is used to calculate the intermediate centrality distribution matrix of the subnet. Finally, traction control nodes are selected according to the distribution of betweenness centrality. We not only preserve the importance of the individual in the system, but also introduce the influence of the importance of the individual in the neighborhood. Comparative experiments show that the betweenness centrality algorithm can greatly enhance the robustness of multi-agent systems and improve the convergence speed of the system.