Localizing an object accurately with respect to a robot is a key step for autonomous robotic manipulation. In this work, we propose to tackle this task knowing only 3D models of the robot and object in the particular case where the scene is viewed from uncalibrated cameras -a situation which would be typical in an uncontrolled environment, e.g., on a construction site. We demonstrate that this localization can be performed very accurately, with millimetric errors, without using a single real image for training, a strong advantage since acquiring representative training data is a long and expensive process. Our approach relies on a classification Convolutional Neural Network (CNN) trained using hundreds of thousands of synthetically rendered scenes with randomized parameters. To evaluate our approach quantitatively and make it comparable to alternative approaches, we build a new rich dataset of real robot images with accurately localized blocks.

The paper presents new results about the geometry of topological interlocking masonries and some possibilities they present to build without formwork. Construction without the use of formwork may be an important issue concerning both productivity increase and decreasing of waste generated on a construction site. Due to the development of computational design and robotics in the construction industry, it makes sense to (re)explore innovative design and process of complex masonry structures. The design of this kind of masonry is standard for planar structures, and in this paper, a generalization is proposed for the parametric design of curved structures. To achieve this, a criterion for translationally interlocked structure based on quadrilateral meshes is exhibited. The application of this criterion is then extended to masonry structures derived from other patterns. Physical prototypes of topological interlocking masonry are also presented. One of these designs seems to allow construction without formwork.

No abstract