Robots could soon be packing up fruit and vegetable orders at Ocado after the UK online supermarket helped develop 'soft hands' that can adapt to the shape of fragile objects to handle them without the risk of damage.
The technology is the result of four years of research conducted by the Horizon 2020-funded SOMA (Soft Manipulation) consortium of researchers, academics, and scientists. Ocado provided the team with a use case for testing their robotic manipulation capabilities, which the supermarket believed could end up on its production lines.
Ocado aims to eventually pick up and pack all of the 55,000 products on its shelves. It will soon launch a system that uses a suction cup to handle more durable objects, but the system will struggle to handle more sensitive and unpredictable items.
"When it comes to classes of objects like fruit and vegetables, we could see that we were going to struggle to find anything that was an off-the-shelf solution, the reason being that with these kinds of items they're never the same shape twice," Dr Graham Deacon, Robotics Research Team Leader at Ocado Technology, tells Computerworld UK.
"You can't, for instance, get a CAD [computer-aided design] model of a bunch of bananas. They're easily damaged. The compliant nature of the hands in the SOMA project meant that we were less likely to impart any damage on the object. And the compliance means that the hands would shape themselves to the object so we didn't need to have accurate geometric models of the objects, the hands would just adapt themselves to whatever shape the objects were."
Environmental adaptation
Robotic arms typically grasp objects with rigid hands and navigate the surrounding environment as an obstacle.
The SOMA consortium took a different approach. Instead of avoiding an object's surroundings, the researchers used the principle of "environmental constraint exploitation" to create an orchestrated interaction between the robot, the object, and the environment, a method inspired by the way humans use their hands.
Deacon illustrates the difference with the example of picking up a credit card from a table. Robotic arms would typically use a navigation system known as dead reckoning, which would rely on sensors to estimate the position of the card. If their estimates are fractionally off, they will miss the card or scratch the table. A human hand has no need for such precision as it can simply slide the card to the table's edge, place a finger underneath and pick it up.
SOMA applied the same strategy to its robotic arms, through a planning system that uses a 3D camera to identify opportunities for environmental manipulation in an individual setting. The pneumatic hand might then choose to drag an object to the surface edge, grab it with a simple top-down grasp, or pin it to a wall to help pick it up, depending on which action has the best chance of success. The fingers will then sequentially close around the object until it obstructs their motion, adapting to the shape that they need to pick up.
"Traditionally, robots move through free space and want to avoid everything apart from the object they're trying to pick up," says Deacon. "They'll pick it up, transport it, and place it without touching anything else in the environment. The SOMA philosophy was about the fact that you can actually use the environment to your advantage and exploit it to improve the robustness of manipulation strategies."
From research to production
Ocado joined SOMA after Deacon was told about the project by an old acquaintance from the Istituto Italiano di Tecnologia (IIT), a scientific research centre based in Genoa, Italy, who thought the Ocado use case would provide a useful test of the technology. Ocado soon agreed to join the consortium, whose other members included the Technical University of Berlin, the University of Pisa, the German Aerospace Centre, and the Institute of Science and Technology Austria.
SOMA appealed to Ocado due to its concept for the hands, the collection of leading robotic researchers who would develop them, and the support of EU funding.
"It really de-risked the whole thing for us," says Deacon. "It was difficult to see how we were going to lose out on the situation, and it helped us do a lot. It accelerated our ability to do a lot more exploration. We've learned a lot along the way and we're going to use that going forward in our production work."
SOMA is one of a number of robotics projects in development at Ocado. Others include the aforementioned suction system, an automated warehouse where over 1,000 robots swarm across a grid the size of several football pitches, and the SecondHands project, another Horizon 2020-funded programme, which aims to create a humanoid on omnidirectional wheels that can assist maintenance technicians.
Ocado aims to add SOMA to its production lines in three to five years but will first explore whether it could benefit from further capabilities.
"We're looking at extending this to being beyond just the hands, but also maybe the arms as well," says Deacon. "We want the whole system to be able to respond to what it experiences when it's in contact with the environment."