Versatile robots need grippers which are capable of both grasping objects forcefully, yet also handling sensitive items with care and delicacy. To that end, researchers are developing grippers which increasingly emulate the human hand not only in terms of shape, but also in their sense of touch.
The hand is a real miracle of nature. Its 27 bones, 28 joints, 33 muscles and thousands of sensory cells enable it to lift heavy loads as safely as raw eggs. Scientists and engineers have for years been striving to imbue robots’ gripping tools with the flexibility and versatility of the human hand. The need for such versatile gripping tools is becoming even more pressing due to the trend towards flexible production processes within the context of Industry 4.0 and the expansion of robot applications into the wide-ranging gripping tasks of everyday life outside of the factory.
And indeed, more and more solutions are being developed which are capable of at least copying the basic skills of the human hand. The key factors in achieving this are carefully applied gripping force and a sense of touch. The German company Schunk, for example, has already brought to market a gripper which uses specially devised gripping strategies and force measuring jaws in its fingers to adapt its responses in real time based on whether it is grasping an item for processing or possibly a human hand. But the Schunk gripper only has two fingers, meaning complex movements such as rotating or rolling a grasped object are not possible.
Robotic hands are learning to grasp correctly
“Hand manipulation is one of the hardest problems that roboticists have to solve,” says Vikash Kumar, a doctoral student at the University of Washington. “A lot of robots today have pretty capable arms but the hand is as simple as a suction cup or maybe a claw or a gripper.” Kumar is part of a team of computer scientists and engineers seeking to find solutions to these challenges. They have developed a robotic hand, with five fingers, 40 tendons, 24 joints and more than 130 sensors, which is able to execute human-like gripping movements. The scientists have provided this skilful hand with machine-learning algorithms. “There are a lot of chaotic things going on and collisions happening when you touch an object with different fingers, which is difficult for control algorithms to deal with,” comments Sergey Levine, assistant professor at the University of Washington. So the scientists devised an algorithm that models the highly complex behaviour of the five fingers on a hand and plans movements in order to achieve different results. During each manipulation task, the system uses its sensors and motion-capture cameras to collect data on the movements executed, which can then be continuously improved through machine-learning algorithms.
Handling even unknown objects safely
The system enables even complex gripping tasks to be realised. But what if the hand needs to pick up delicate objects such as a raw egg? That task was the focus of a gripper designed by the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. It consists of three soft silicone fingers which wrap around delicate objects, grasping them very safely and protectively. If it encounters a hard or unusually shaped object, it grips it using only its finger tips, providing for greater precision with increased point pressure. The gripper uses sensors to identify the shape of an object and so decide how to handle it. The robot’s processor unit is able to identify the object based on just three data points. To do so, the system compares the data against past gripping tasks. “If we want robots in human-centred environments, they need to be more adaptive and able to interact with objects whose shape and placement are not precisely known,” explains Daniela Rus, Head of the Distributed Robotics Lab at the CSAIL. “Our dream is to develop a robot that, like a human, can approach an unknown object, big or small, determine its approximate shape and size, and figure out how to interface with it in one seamless motion.”
Hairy sensors
This also requires new sensors which not only register the shape of an object, but can also detect whether it is soft, or whether it starts to slip because its grip is not firm enough. As one solution, the Tacterion corporation is developing tactile sensors in the form of a thin polymer-based artificial skin. Similar to the nerve cells in the human skin, its touch-sensitive sensors enable it to “feel” even the lightest pressure and process the corresponding data. Chinese researchers Rongguo Wang and Lifeng Hao at the Harbin Institute of Technology are working on the development of ultra-fine hair sensors for an electronic skin. The sensors are cobalt-based micro-wires encased in a thin layer of glass and embedded in a robust, rubber-like skin sensor. The hairy robot skin is able to detect a fly landing on it, the feel of a light breeze, or a five kilogram weight. The innovative sensor can also sense when an object it is gripping is starting to slip. “The sensor has a number of extraordinary capabilities, including the ability to detect air draughts, and to characterise material properties, as well as offering excellent robustness against damage,” the researchers claim.
(Picture Credits: Istockphoto: CSA Images)