Recently, the industrial revolution has led to the emergence of a new generation of robotics called as modular robotic system (MRS). However, such new technology faces technical and practical challenges, in particular the self-reconfiguration. Subsequently, converting from one morphology to another is a complicated task for MRS that may take a long time because of the huge number of communications needed among the modules. In this paper, we propose an efficient and robust cluster-based planning, called RUN, for fast self-reconfigurable modular robots. RUN works on two stages and aims to group the modules into clusters in order to reduce the communication overhead between them and offers a fast reconfiguration process for the MRS. The first stage selects a set of modules, called cliques, and then, it divides the modules into clusters based on the shortest paths between modules and cliques. The second stage introduces two communication algorithms: the inter-module algorithm that allows an efficient communication between the cliques of the clusters, and the intra-module algorithm that reduces the communication number between the modules of the same clusters. We show the efficiency of RUN in terms of communication reduction and fast reconfiguration process, through simulations on Roombots compared to other exiting techniques.