What changes when a robot “hand” is no longer tied to the arm it works for? A group of researchers with the names Xiao Gao and Aude Billard has presented a solution to this problem by designing a detachable and reversible robotic hand that can crawl, grasp, lift, and then reattach itself to its base on an arm without having to let go of the object that it has picked up. The solution presented by the group has been published in Nature Communications and presents a combination of manipulation and locomotion as one problem instead of the two problems presented in the previous solution.
The first mechanical difference is symmetry. Most robotic hands are based on the human hand, with one palm side facing out for interaction and an opposing back that is mostly dead space. This kind of design is very helpful when doing a lot of pick-and-place tasks, but it is a pain when doing tasks in tight spaces and with multiple objects, especially when the object is “behind” the grasp and the wrist is not very flexible. Billard encapsulated the reason for their design choice in one sentence: “We can easily see the limitations of the human hand when attempting to reach objects underneath furniture or behind shelves.”
The two prototypes that will be used as the anchors of the study are the five-fingered and six-fingered prototypes, which are designed around a palm with a diameter of 16 centimeters and are powered by small electric motors and made of 3D-printed joints that are light in weight. The fingers bend and extend in a way that is very much like the human hand, but with the ability to flip the working surface of the hand in a bidirectional manner. The fingertips are covered with a soft silicone layer.
The detach-and-return cycle is no gimmick; it is the enabler of reach. Instead of having to send the entire mobile robot to retrieve a dropped part, the hand can simply detach from the arm, crawl to the target, and retrieve it. A snap-and-lock docking system with magnetic registration and a motorized locking bolt is the mechanical handshake required for docking. The end result is a manipulator system that extends the robot’s working envelope into the corners, shelves, and crevices that are impossible for rigid-link robots.
In the demonstrations, the hand could reproduce 33 human grasp patterns with the five-finger hand and lift 2 kilograms of weight. It also demonstrated the capability to do sequential retrieval and return, re-grasping while maintaining a steady grasp of up to three objects simultaneously; these included common objects such as a cardboard tube, a rubber ball, a marker, and a can. The six-finger model demonstrated the importance of symmetry and additional fingers, as it could do tasks that require the use of two hands, supporting one hand while the other does the task. However, not all trade-offs are excluded.
As quoted in the discussion of this research by external roboticists, the presence of a finger that can bend in multiple ways may cause stiffness and force transmission issues relative to a finger designed for bending in one way. The focus of this research is still on grasp stability, mobility, and reconfiguration, and not on in-hand manipulation. However, perhaps the most critical role of the hand is more conceptual in nature. It considers “where the hand is” and “what the hand can hold” as strongly coupled design variables. In industrial service robotics, inspection robots, and other systems that have to operate around geometry rather than in front of it, the crawling and reversible hand allows the end-effector to be a short-range robot on its own.