An important logistics application of robotics involves manipulators thatpick-and-place objects placed in warehouse shelves. A critical aspect of thistask corre- sponds to detecting the pose of a known object in the shelf usingvisual data. Solving this problem can be assisted by the use of an RGB-Dsensor, which also provides depth information beyond visual data. Nevertheless,it remains a challenging problem since multiple issues need to be addressed,such as low illumination inside shelves, clutter, texture-less and reflectiveobjects as well as the limitations of depth sensors. This paper provides a newrich data set for advancing the state-of-the-art in RGBD- based 3D object poseestimation, which is focused on the challenges that arise when solvingwarehouse pick- and-place tasks. The publicly available data set includesthousands of images and corresponding ground truth data for the objects usedduring the first Amazon Picking Challenge at different poses and clutterconditions. Each image is accompanied with ground truth information to assistin the evaluation of algorithms for object detection. To show the utility ofthe data set, a recent algorithm for RGBD-based pose estimation is evaluated inthis paper. Based on the measured performance of the algorithm on the data set,various modifications and improvements are applied to increase the accuracy ofdetection. These steps can be easily applied to a variety of differentmethodologies for object pose detection and improve performance in the domainof warehouse pick-and-place.
展开▼