Grasping of Unknown Objects with an Autonomous Manipulator: State of the Art, Problems and Prospects

To fulfill the practical needs of modern robotics, it is necessary to develop approaches for grasping unknown objects, since in the real world the robot faces a large variety of them. Approaches that imply the availability of complete information about the objects of the working area (3D model, weight and size characteristics) are not practical and can only be used in controlled conditions, such as working on a conveyor with standard details. Therefore, the scientific community and a number of industries are interested in research methods that increase the robot’s ability to adapt to new, unfamiliar conditions. This article presents main problems and research directions in the field of visual scene perception and grasping unknown objects by a manipulative robot. We discuss the differences in existing approaches according to various criteria, as well as advantages and disadvantages of existing solutions. The article may be useful to get acquainted with the subject area.