The study of load-induced cascading failures in interdependent networks is of great realistic significance, which can provide valuable reference for designing high robust interdependent network or improving their robustness. In this paper, we establish a cascading model for a double layer interdependent network, and study the effects of the contributions of inter-degree and intra-degree to the loads, the coupling, and the intra-node linking similarity on the cascading failure in the interdependent network. Our studies show that when the contributions of inter-degree and intra-degree to the loads attain some values, the interdependent network reaches the highest robustness against cascading failures. As a notable feature for the interdependent network that is different from an isolated network, the coupling must have a significant influence on cascading failure in the interdependent network. In order to reach higher robustness, we suggest that the disassortative coupling be used and the inter-degree be made as homogeneous as possible under condition that a larger average inter-degree is adopted. In addition, we find that it is contrary to the case of neglecting loads that when the Pearson correlation coefficient for measuring the intra-layer degree-degree relation is larger, the interdependent network is more robust against cascading failures.