Passive 3D printed robot hand can identify and grip objects

Human hands can be very dexterous tools and though the level of skill they wield can vary from person to person, from a very young age people have an innate ability to identify how much pressure to use when gripping an object. However, teaching the principles of pressure and dexterity to a robot can prove challenging.

The robotic hand created by Cambridge University’s Department of Engineering utilises wrist motion to elicit a range of movements. Like the claw machine you might find in a children’s arcade, the hand is able to move over objects, bear down on them, and grip them but, unlike the claw machine, this hand doesn’t drop the object it has picked up.

This is because tactile sensors are placed in the fingers of the anthropomorphic hand, which allows the robot to gather information on what kind of object it is picking up, and how much pressure it needs to successfully hold onto the item without dropping or destroying it.

Co-Author of the paper, Dr Thomas George-Thuruthel, who is now based at University College London (UCL) East, comments: “In earlier experiments, our lab has shown that it’s possible to get a significant range of motion in a robot hand just by moving the wrist. We wanted to see whether a robot hand based on passive movement could not only grasp objects, but it would be able to predict whether it was going to drop the objects or not, and adapt accordingly.”

This is not the first-time researchers from Cambridge University have combined hands, robotics, and 3D printing, with complex movement and control. In 2018, researchers used both soft and rigid materials, which were intended to replicate bones and ligaments, to train the robot hand to use passive movements to play a piano.

By not creating a fully actuated robot hand with individual finger motors, the team are also able to save significant amounts of energy, keep costs low and have easier control of the hand. It also allows the hand to become adept at working passively so that when the team do eventually add any additional functions, such as motors or cameras, it will be able to perform far more complex tasks efficiently.

More than 1,200 tests were carried out, and the robot’s ability to grasp objects without dropping them was observed. Using a pre-defined action obtained through human demonstrations, it was initially trained to grasp small 3D printed balls.

By a process of trial and error, the robot was able to learn which grip would be effective for various objects, and it successfully grasped 11 of 14 items which included a computer mouse, a roll of bubble wrap and a peach.

“This kind of hand has a bit of springiness to it: it can pick things up by itself without any actuation of the fingers. The tactile sensors give the robot a sense of how well the grip is going, so it knows when it’s starting to slip. This helps it to predict when things will fail,” comments first author Dr Kieran Gilday, who is now based at EPFL in Lausanne, Switzerland.

Gilday continues: “The robot learns that a combination of a particular motion and a particular set of sensor data will lead to failure, which makes it a customisable solution. The hand is very simple, but it can pick up a lot of objects with the same strategy.”

Other robotic arms have been successful in identifying, gripping, and moving objects on an industrial level, such as Amazon’s robot ‘Sparrow’. However, it is an interesting and exciting step forwards for passive movement robotics and how they might eventually mimic the natural movement and delicacy of handling items – just like a real human hand.

But, if robots do eventually replace the need for humans to perform tasks that require sensitive and tactile actions, it begs the question: what will they be capable of performing in the future?