Abstract: The stability of nonlinear large scale systems containing contrained uncertain parameters is studied in this paper. This kind of stability is called robust stability when the vectors of the parameters are constrained. The large scale system can be reduced to a linear system by aggregation linearization via Liapunov functions in the finite region of the state space containing the origin. This linear system obtained is not unique due to the flexibility occuring in linearization. It is shown that the hierarchical structure of the large scale system estabilished by graph theoretic method can simplify the finite region stability criteria and help in choosing local feedbacks for stabilizing the large scale system.