In a vast Amazon warehouse on the outskirts of Tracy, central California, flat rectangular robots wait to be handed packages by robotic arms. The robots then scuttle away to hurl the parcels down chutes, for human workers to load into delivery trucks.
“Each one is their own beast,” says Jonathan Ramey, of the robots he maintains. “They work through thousands of packages an hour and they don’t let up.”
Robots like this are playing a bigger role in warehouse networks, with the incorporation of artificial intelligence enabling major companies to scale these technologies as they seek to beat rivals on convenience.
Amazon is testing more advanced systems as part of its drive to increase efficiency in sorting and fulfilment centres. Investments in robotics and AI are part of the ecommerce giant’s $100bn planned capital expenditure this year.
Reaching this point has taken Amazon more than a decade.
Why did Amazon start using robots in its warehouses?
The Seattle-based group is now one of the world’s largest users of industrial robots and its appetite for the technology has shown no signs of abating as it seeks to meet chief executive Andy Jassy’s ambition to drastically reduce delivery times.
“Our robotics touch billions and billions of packages that we ship each year, says Tye Brady, chief technologist at Amazon Robotics. It is a “great flywheel” that increases productivity while reducing the strain and monotony for human workers, he adds.
Amazon has expanded its robotics operation dramatically since it acquired start-up Kiva systems for $775mn in 2012. Since rebranded Amazon Robotics, the division has rolled out more than 750,000 mobile devices and tens of thousands more robotic arms and autonomous systems.
The company has more than half a dozen types of robotic system in its logistics operation, spanning robotic arms through to heavy lift units and sorting machines.
What can Amazon’s robots do?
The first mobile robot deployed in Amazon’s warehouses was the Kiva robot, which lifts and moves shelving units from below. This system evolved into the so-called Hercules and Titan robots, some of which can lift over a tonne.
It has also rolled out robotic arms such as Cardinal that pair suction cups with a set of sensors and cameras to determine the amount of pressure needed to lift and move an item safely without causing damage. Other machines help with moving, packaging and sorting items as efficiently as possible.
Amazon’s robots
Amazon’s Sequoia system takes plastic crates containing inventory from shelves and passes them to employees, making it quicker for them to grab items
Smaller “drive units” carry packages across the warehouse to designated chutes, while also being able to lift shelving units and take them to employees.
Robotic arms such as Robin, Sparrow and Cardinal select, pick up and sort items, either for employees to box up or to be loaded on to delivery trucks.
Amazon’s Proteus drive unit (pictured) has been trained using artificial intelligence and deploys computer vision to freely navigate spaces alongside employees
Bespoke software and tools developed by Amazon are built on top of hardware manufactured by companies including ABB in Switzerland and US-based Fanuc.
Robotics has been implemented in each stage of Amazon’s fulfilment process, though the company maintains that human workers continue to have a role in its operations.
How are workers affected?
Eva Ponce, a director at the MIT Center for Transportation & Logistics, says the technology is replacing some jobs — particularly more manual tasks including heavy lifting — but has also offered new opportunities within warehouse operations.
“Companies are investing more in upskilling associates . . . ensuring they’re ready for a new style of work,” Ponce adds. “More complex tasks will still need to be done by people.”
Amazon workers still have to handle awkwardly shaped packages, those that have been improperly labelled and deal with returned items that need to be inspected for damage.
The group says it has spent about $1.2bn since 2019 training up workers for maintenance roles. It has pushed back against the notion that it is seeking to replace all workers with machines, noting it employs more than a million more people than it did when it bought Kiva in 2012.
Yet working conditions within its warehouse operations are a constant source of tension with labour organisations, who warn that robotics-enabled warehouses increase the pace of daily work and drive injury by pushing workers beyond the limits of what is safe.
A US Senate committee report published in December showed that Amazon warehouses recorded a 30 per cent higher incidence of injury than the industry average.
Amazon said that robotics and AI “play an important part in supporting our company-wide safety objectives” and that it expected the technologies to improve safety through better ergonomics and less heavy lifting.
How much more efficient are warehouse robots?
Amazon’s investment in robotics has been mainly motivated by a drive to lower costs, using any efficiency gains to keep prices down, while also meeting ambitions to deliver some items within hours.
Jassy told investors during a recent earnings call that the business would continue to invest in robotics and automation.
A warehouse opened by the company in Shreveport, Louisiana, last year has 10 times as many robotic pieces of equipment as previous versions. It estimates that the investment has led to a 25 per cent reduction in the cost of fulfilling orders.
More investment in this type of warehousing operation will generate around $10bn in annual savings for the ecommerce group by 2030, analysts at Morgan Stanley estimate.
How is this changing ecommerce?
Amazon’s rivals, including Walmart, are also making significant investments in robotics and automation, with Ponce noting that the desire to be more efficient and boost productivity has accelerated amid labour shortages.
“These technologies are disruptive,” Ponce says. “The warehouse of the future is a combination of robotics, sensors and computer vision.”
Brady says Amazon is making further inroads using artificial intelligence. Its Proteus unit uses “computer vision” — deploying AI to enable robotic systems to process their field of vision in real time — to help it navigate spaces alongside human workers.
Proteus has enabled the company to shrink the size of warehouses, while investments alongside chipmaker Nvidia to create “digital twins” — simulated versions of its warehouses — allow it to train autonomous robots at an unprecedented rate.
“This has completely changed how we do business . . . [And] it’s just the beginning,” says Brady.
