In this paper, we present a modular method of environment perception for a mobile robot by local mapping. This map building process is done in order to authorize the motions of a mobile robot in unknown indoor environment. To validate it, we have implemented a reactive navigation on a mobile robot which perceives its near environment with two range finders. The map building process is the cluster of several functions: sort of measures in subsets, model of these subsets, elaboration of a map, and updating it with previous measures. Furthermore this process must be quick and sufficiently robust to deliver in real time a safe map to the robot.