Novel Pose Estimation System for Precise Robotic Manipulation at CERN

The “mission” of tele-robotics at CERN may be resumed in the following: Ensuring safety of personnel improving availability of CERN’s accelerators. The robots that are being developed at CERN should offer visual capacities, among them the capacity to estimate the 6D pose of an object. To know the position and rotation of an object could be used in augmented reality for a better information for the tele-operator and for the realization of semi-autonomous tasks. The goal of this project is to develop a novel and reliable algorithm to estimate the 6D pose, using computer vision, of objects that are going to be telemanipulated by robots. The novel algorithms should work consistently in the CERN unstructured and harsh environments, presenting several constraints like variable luminosity, difficult accessibility and light reflections. In this project, an algorithm has been developed to detect the position and rotation of objects in the LHC tunnel using 3D cameras. The objects selected for this project correspond to a collimator. The collimators are of vital importance for the correct operation of the LHC and therefore, have great importance in technical inspections. In addition, they are one of the hot spots of radiation in the LHC. The algorithm has been developed using Point Cloud Library to manage the point cloud created with an RGBD Camera. Due to reflections on the target material, the point cloud contains partial information of the desired object. Once the position and rotation of the object is known, it is possible to reconstruct the lost information using CAD models.