Robotic gripping and dexterous manipulation

Since the birth of the first industrial robot in the early 1960s, robotics has replaced humans for many often tedious and repetitive tasks in the world of industry. To meet these challenges, industrial robots have become specialized. They have been conceived and designed with the task in mind, and the same applies to the grippers or end units that equip these robots. In the early 1980s, the temptation to provide the robot with a universally-applicable end-member led to the development of the first robotic hands, which marked the history of robotic gripping. Today, the emergence of collaborative robotics and cobots calls for the development of new-generation, flexible grippers, equipped not only with adaptive grasping capabilities, but also with dexterous manipulation capabilities.

The model of the human hand, with its functional richness and infinite capacities, is an unrivalled reference in terms of dexterous prehension and manipulation. Replicating these functions in a robotic system remains one of the most complex problems to solve.

The gripping of parts and/or objects requires, depending on their nature, the use of adapted mechanical systems, ranging from dedicated industrial grippers to highly dexterous robotic hands. The complexity of the gripper will thus increase according to the desired level of flexibility, from gripping objects of various shapes to manipulating them inside the gripper. The integration of sensors is necessary for several reasons: locating the object gripping surfaces, determining the gripping configuration, controlling the gripping effort and, more generally, controlling the actions performed by the gripper.

The low-level control of these systems must ensure that the gripper's movements and the interaction force are controlled to capture objects and maintain a given configuration.

For more complex functions, such as manipulating an object with a fingertip, more sophisticated commands are used to plan and control the coordinated movement of the fingers as they interact with the object. This coordinated control of the fingers is in itself an exercise in collaboration between the elementary robots that are the fingers of a robotic gripper with a high degree of actuation.

The dynamic interaction of a cobot equipped with a gripper with the environment, and in particular with humans, increases this complexity. The ability to adapt and perceive the environment must therefore be integrated into a higher-level control system capable of responding in real time, based in particular on learning and the implementation of reactive planning strategies.