Simply: grasp!

点击翻译按钮获取中文摘要

Grasping is a grand challenge in robotics. How to pick up objects made of different materials and shapes is of the utmost importance to truly unveil the potential of robotics for assisting humans in work and daily living activities. While many would like domestic robots with advanced pick-and-place capabilities to, for instance, unload their dishwasher, industry would highly benefit from innovative grasping solutions able to automatically pick and sort items of different shapes and materials in their warehouses, plants and stores. There is a wealth of prolific research on vision-based systems for robotic grasping. A primary focus has been on identifying the position and orientation of the target, i.e. the object to be grasped. Research also focused on finding optimal grasping locations on a given target based on the specific constraints of current grippers. Research additionally investigated how to best identify optimal manipulator trajectories to precisely reorient grippers, as current commercially available grippers require to be precisely oriented to grasp. While much progress has been made on the above-mentioned aspects, robotic grasping is often still unsuccessful in practice. In fact, currently available commercial grippers often simply cannot grasp targets with different shapes and made of different materials. Additionally, vision-based systems may fail to provide a target's precise orientation in real-world environments, such as a house, making the currently needed process of deciding how to best orient the gripper very challenging. This research program therefore aims at generating innovative grasping solutions which require minimal reorientation to grasp targets and are able to grasp most targets of different shapes and materials, even when little information about them (e.g. orientation) is available. This research has the potential to contribute to the wider adoption of robotics in our society and assist with economic growth. A grasping solution enabling most objects to be grasped has the potential to make a transformative change in robotic grasping. Results of this research program, when paired with state-of-the-art vision-guided dexterous robotic manipulators, could help e-commerce and other businesses to dramatically improve plant automatization and economic growth of our industry. Grasping solutions that are also effective when little information about the target is available have the potential to enable new industries to use robots in their facilities, even in poor lighting conditions and in the presence of occlusions. Additionally, grasping solutions requiring only minimal reorientation will enable the extensive use of affordable manipulators with few degrees of freedom guided by inexpensive cameras. These low-cost robotic systems are expected to make a strong impact in our society, as cost is a primary barrier to entry in many industrial applications. They also have a strong potential to find future daily use in our homes.