Multi-Modal RGBD Sensors for Object Grasping and Manipulation

RGBD sensors, such as the Microsoft Xbox Kinect [1] are types of multi-modal perceptual sensors that have appeared in recent years. RGBD sensors have become standard perceptual tools for robots as they provide a unique multi-modal approach to perception. A vital pre-cursing challenge in object grasping and manipulation is object pose recognition. A robot must identify the pose (i.e. orientation) of an object in order to perform grasping/manipulation tasks accurately. In this work we focus on combining multi-modal RGBD data to reduce uncertainty and hence solve the problem of object pose recognition more accurately. We experiment on an RGBD dataset and show that our approach has a significant improvement of more than 20% over