[ad_1]
The most striking sight in an Amazon fulfilment centre is the dozens of 3 metre tall yellow shelving units that pirouette and move about each other in the facility’s centre.
These pods were carried about on small square blue robots, and moved to allow human pickers on the sidelines to grab products from within them to fill orders.
This coordinated dance of robotic shelves is visually impressive, but it is the efficiencies being achieved from artificial intelligence (AI) that reveal how stockrooms and distribution centres will work in the future, and what companies will have to do to keep up.
Shaown Nandi is director of tech for strategic accounts at Amazon Web Services (AWS).
Although AWS was a separate company, it was responsible for creating the AI systems in operation within Amazon’s fulfilment centres.
The list of how AI was used is long. AIs decide whether certain items should go in a padded envelope or box, and exactly which size of box should be used.
“We’ve been very focused on that for a long time and since 2015, we’ve reduced the weight of our outbound packaging by almost a third, which is a ton of waste removed,” Nandi said.
The same AI system, built on AWS’ SageMaker product, also recorded whenever a product was damaged on delivery.
If it was, the AI learnt, and decided if damage was common enough to make a change.
Other AIs were able to look at pictures taken of product bins, and decide whether they were full, or could fit more.
These were all tasks, Nandi said, which no human enjoyed doing.
AWS’ Monitron was another system in use within fulfilment centres, which interpreted data from sensors within industrial equipment, like the long conveyor belts.
Supplied
The cherry-pickers in Amazon’s centres are programmed to be unable to move outside their designated areas.
Things like the level of vibration were analysed by an AI, which looked for patterns or clues that suggest there was a problem, or likely to be one in the near future.
The sign could be as small as vibrations caused by a bad bearing.
“Everyone has different types of equipment that need to be monitored, but vibration is a common thing. What you do with that vibration data is going to be quite different,” Nandi said.
AI was also being incorporated into supply chain management.
Amazon was currently building a system that would allow for predictive inventory management. Essentially building an AI that could guess what people would buy, when they would buy it, and how many they would buy, in order to get those products in ahead of time.
When asked whether this type of tech would become a necessity within other company’s distribution centres in order for them to continue to compete, Nandi preferred to frame them as opportunities, rather than necessities.
The same AWS products could be used by Amazon’s rivals, and adapted to their needs, he said.
“What you’ll find is the whole purpose of AWS is to take these building blocks and make them available to customers who want to build with them,” he said.
“I think you’ll see some elements of machine learning or AI fall into almost any company’s operations, certainly in logistics and distribution, and there are a lot of different ways to do it.”
How fulfilment centres work
Deliveries arrived on large trailers and were sorted into the robotic pods by either a human, or a robotic arm.
If it’s a human, bright lights illuminate appropriate pigeonholes to put each product in, informed by a computer that monitors how full they were.
A database records which pigeonhole in the pods on the unit an item was placed into.
When that item was ordered, the system called that roving pod to the front, directing all others out of its way, where a human picker was shown the right pigeonhole and product to pick.
Pickers passed the product to packagers, who were advised exactly which size of bag or box to use, and then it was onto a 100 metre conveyor belt, where packages were sorted into location, and passed on to delivery drivers.
These drivers were most often Flex drivers, contractors who worked almost like Uber drivers, taking on shifts and delivering parcels in their own cars.
It is this system that allowed one centre in New Jersey, near New York, to facilitate 30,000 deliveries on any given day, and on busy days, to fulfil over 110,000 orders.
Supplied
Amazon’s Prime Days event sees purchases jump to almost four times the usual rate.
The site lead of the New Jersey facility, Drew McCrossan, is an ex-navy officer of 21 years, flying helicopters and drones.
He said Amazon attracted a good number of veterans, partly because the systems were very ordered and regimented, and because there was a “mission accomplished” mindset to getting orders out on time, which at this facility meant for same-day delivery.
“There’s a high level of integration between our human associates that are working in different paths and all the technology we have here,” he said.
“We have algorithms that are working back-end that you don’t always see.”
[ad_2]