Last updated May 20, 2019 at 11:04 am
Scientists are helping robots hold and pass objects more naturally by studying how humans do the seemingly simple but subtle task.
Scientists have analysed over 5,000 one-handed human grasps and handovers of objects in an attempt to help robots grasp objects better.
While grasping an object, such as a cup or ball, may sound like a simple skill, most robots are currently unable to reliably perform the action.
By analysing people, the researchers mapped out how we choose to grasp and our hand placement on objects when we need to pick them up and hand them to a receiver, and how that second person then takes the object.
The research, led by Francesca Cini from The BioRobotics Institute of Scuola Superiore Sant’Anna and Valerio Ortenzi from the Australian Centre for Robotic Vision at the Queensland University of Technology and published in Science Robotics, involved objects as diverse shaped as a pen, a screw driver, a bottle, a book, and a toy monkey.
Getting a grip
Humans intuitively grasp and exchange objects using unconscious and intuitive signals and actions, such as leaving handles or space available for the receiver to grab, allowing the natural and smooth completion of what is essentially a collaborative task.
“A handover is a perfect example where little adjustments are performed to best achieve the shared goal to safely pass an object from one person to the other,” says Ortenzi.
However, with further integration of robots into households, hospitals and industry on the horizon, robot actions have previously been found to be unsafe or unnatural when it did not deliver the appropriate cues.
Choosing the right grasp in this manner is not as simple for robots, especially since grasp choices depend on the situation – involving a variety of factors without one single factor playing the most important role.
Passing on the human technique
The researchers evaluated 5,202 different grasps of various objects, involving 17 pairs of people passing 17 different objects back and forth.
The passer first grasped the objects from a table and performed two different tasks with each of the objects, then handed the objects over to a receiver who subsequently performed the same two tasks.
Example of trials with the views of the object tracking (the trials shown in the video were recorded to explain the experimental procedure and were not included in the analysis of grasp type and location). Credit: Cini et al., Sci. Robot. 4, eaau9757 (2019)
The researchers found that across various object handovers, passers favoured precision grasps, in which only the tips and upper halves of fingers are used (73% of cases).
As well as leaving the “handles” free for the receiver to grab, humans also tend to leave the purposive part of objects unobstructed to the receivers, allowing the receiver to comfortably grasp the object and quickly perform the next task with maximum dexterity, the researchers say.
These behaviours are generally unconscious to humans, however they are patterns that our brains have learned over time through repetition and routine. Learning this grasping and manipulation process is subtle and elusive for robots,” conclude the researchers.
To further this work, the QUT team has recently been awarded a US$70,000 research grant from online giant Amazon.