Researchers in the US claim to have developed a robotic system that can not only pick and pack goods in warehouses, but is able to identify objects and treat them accordingly - prevent clumsy robot hands from squashing tomatoes, for example.
The development by scientists at MIT and Princeton University is similar to a number that UK online supermarket Ocado has been working on, including a robotic arm capable of telling the difference between, for example, bags of tomatoes, boxes of eggs and bags of potatoes, and treating them accordingly.
The MIT robotic device, which the scientists described a "pick and place system", comes with an industrial robotic arm that is kitted out with a gripper and suction cup. It has what the researchers have lavelled "a grasping algorithm" to enable it to pick up any object, based on the properties of the object.
According to the academics, the system can also work out the best way to grab a product, even if it has not done so in the past. It can lift items out of bins, too, enabling it to pick and pack small items.
The robot is assisted by a string of cameras, which capture the object in various angles so it can work out what it is picking it up. It does this by comparing the shots to an image library.
The scientists said that the robot uses a "grasp-first-then-recognise" workflow. Alberto Rodriguez, a professor at at MIT, claimed that the robot is more effective when compared to similar technologies.
"This can be applied to warehouse sorting, but also may be used to pick things from your kitchen cabinet or clear debris after an accident," he said.
Ocado Technology, the technology development arm of Ocado, has already shown off pick-and-place robot technologies. It has also unveiled a robot capable of helping out maintenance technicians, developed as part of the EU-funded SecondHands project.
However, the MIT scientists believe that "most existing systems are typically designed to function only in tightly controlled environments".
Their robot, they claim, is "more flexible, adaptable and intelligent pickers" and can be used in a variety of settings. The system not only grasps objects, but can also recognise and classify them.
The MIT and Princeton research team taught the robotic arm to get items from a cluttered bin. They said it does this by choosing from four grasping behaviours.
"We developed a system where, just by looking at a tote filled with objects, the robot knew how to predict which ones were graspable or suctionable, and which configuration of these picking behaviours was likely to be successful," said Rodriguez.
He added: "We're comparing things that, for humans, may be very easy to identify as the same, but in reality, as pixels, they could look significantly different.
"We make sure that this algorithm gets it right for these training examples. Then the hope is that we've given it enough training examples that, when we give it a new object, it will also predict the correct label."
The researchers also suggested that their robot technology could also be used in households and work alongside people in other workplaces.
They envisage the robot completing household chores, but said it that could also potentially organise products in shops and clear debris in disaster zones.
Parker Solar Probe, TESS and GOLD missions will deliver exciting data, claims NASA
But deep learning pulls ahead for complex tasks
Geoengineering on the sea floor near glaciers would form a new ice shelf to prevent melting
Alterations in capillary blood flow can be caused by body position change