Amazon A new warehouse robot has been developed that uses touch to bounce ail around the shelves to find the right product to ship to customers.
The robot is called Vulcan and is a meaningful step in making robots less finger measures compared to humans. Further honing the tactile abilities of robots may allow them to do more achievements and manufacturing work in the coming years.
Aaron Parness, head of Amazon Robot AI, leads Vulcan's development, explains that Touch Sensing helps robots push items onto the shelf and determine the consequences. “When you try to store items in one of the pods, you can’t really complete the task without contacting the other items,” he said.
The Vulcan system consists of a conventional robot arm with a custom scraper-like appendage for poking into the rack, and a suction cup to grab the item to pull it out.
Vulcan has sensors on several of its joints that allow the robot to detect the edges and contours of an object. Machine learning is key to understanding sensor signals and also forms part of the algorithmic loop that controls how a robot takes action, Parness said. “The special seasoning we have is a software explanation of force torque and how we wrap them in a control cycle and into the exercise plan,” he said.
Amazon revealed Vulcan today at a fulfillment center in Hamburg, Germany. The company said the robot is already working at the facility and another plant in Spokane, Washington.
The new robot will work on the same line as the human picker and will grab more items from more items on the shelf, which grab more items from above or lower shelves. The robot believes that the items that cannot be found will be reassigned to human workers.
"Amazon stores many different products in the bin, so it needs to be done a renovation to fill in a specific object to fill in the order," said Ken Goldberg, a robotist at the University of California, Berkeley. "That has been difficult so far, so I'd love to see a new system."
Goldberg said that in recent years, research on robot touch sensing has developed, with many groups engaged in joint and surface sensing. But he added that robots still have a certain way to go before matching the tactile abilities of flesh and blood workers. “Human touch is very sensitive and complex, with a large dynamic range,” Goldberg said. “Robots are developing rapidly, but I was surprised to see human equivalent (skin) sensors for the next five to ten years.”
Even so, Vulcan should help automate more of the work done by humans within Amazon’s vast fulfillment center. In recent years, the company has improved automation through AI-injected robots that can capture and transport packages and boxes. Store and retrieve items from shelves is one of the more challenging tasks a robot needs to do, and it depends to a lot on human labor.
Parness said he would not foresee the robot taking on all the work done inside the Amazon fulfillment center. “We really don’t believe in 100% automation, nor is it lights satisfying,” he said. “We can get to 75% and have the robot work with the staff, and that amount is much more than working alone.”