Abstract. This article presents a method for cooperative reconstruction of three-dimensional scenes using múltiple views. Although the use of múltiple views is one of the most applied in the 3D scenes interpretation, its use with a single mobile robot does not guarantee the perfect localization of the environment, due to odometry errors. Also, the reconstruction of dynamic environments is not allowed for a single robot, which takes views in different instants. This work proposes a system formed by múltiple robots to get a cooperative reconstruction of the scene. The robots decide the best strategy to acquire the image to get the best reconstruction using an objective function defined. In this function, mainly are considered uncertainty of the reconstruction and view points. The result of this optimization is the next position of the robots.