Inside the automated warehouse where robots fill in your grocery stores

It is the lack of noise you notice first. There is no stapler of equipment, engines or gossip from co -workers. Only the home of electronics. For the industrial space, this is frighteningly calm, but it is logical in a building that may exceed the number of robots.

I am in a warehouse – or the Al -Wafa Customer Center (CFC) – run by the OCADO grocery company in Luton, outside London. You may not have heard about Ocado, but may still have handed over your groceries. Technology is online orders for Kroger in 14 US states, Sobyys in Canada, both Morrisons and its UK’s delivery mark, with other customers across Europe and Asia.

Grocery work has tight margins, and online orders more, as stores are forced to have additional costs to capture packing and shipping orders. OCADO, which was launched in 2000, has always been in favor of using automation to reduce these costs.

In the heart of everything “network”. This sprawling allows most of the upper floor of the warehouse, this intersection of the tracks is a fleet of hundreds of huge and huge robots to WhOSH, which is carefully controlled by a central computer to avoid collision, and move the customized baking trays, ready -made meals, and more in any place they need. It is almost completely a conference – to the extent that while I stand from the maintenance corridor, I can’t see one soul regardless of the OCADO employees who guide me that morning. Humans are not required to supervise robots or work alongside them. Even technical support is dealt with remotely, by a team in Bulgaria. At some point, I see the LEDS Green Robot turns amber, indicating a problem. It stops quickly, sits there, orange, for 30 seconds or so, then covers the green and rolls happily again, there is no required personal assistance.

Nothing of this new. In fact, not even the first time freedom Watch it – again in 2018, We visited another CFC in the United KingdomWhen the network was sophisticated. Now it’s old news. It has not been replaced, but in the midst of the upgrade that adds an important element: weapons.

Let’s back down for a moment. Cuban robots on the network do not pack shopping bags for anyone. Until recently, they were only assigned to animated boxes, seizing a box of beans from the waterfall where they are stored, and transferred to another farm, where they decrease alongside a human worker in time so that they can pack a few cans in a person shopping bag. Workers are expected to pack elements in seconds, and the system works because it is very effective. By the time when the employee is ready to pack a component, it is already on their side, and the screen tells them the number of bags in which the boxes are made. Even their instructions are calibrated by the computer to reduce the unnecessary movement that may slow them down. This is a human act, but it has been improved to its limits.

But now there is something new. It floats across the network, high education over the squatting citizens, is a new type of robot. This sitting is in place, and the islands are in the case of the continuous movement around them. But just like people who stand on the floor below, they are busy filling the bags.

The choice of robotic on the network (OGRP), each arm is equipped with a small suction cup at one end. Sixty five of them sit on the network in Loton, with 500 original robots that bring them boxes, others with customer shopping bags, others with groceries are ready for packing, choosing weapons and packing them in bags. Each OGRP arm has a camera to help capture groceries, but it is not designed to identify damaged goods, so they will not discover broken eggs or apples with bruises, giving humans at least some advantages.

In 2024, OGRP More than 30 million requests With the installation of less than 100 cubits, by the end of this year, OCADO expects that there will be approximately 500. James Matthews, OCADO CEO, tells me that weapons are now able to package about 40 percent of OCADO groceries. The company expects to reach a range of 80 percent, partly by introducing a set of new end point attachments to match the current suction cup, from parallel to a soft one. This is not a decade away, either – this is where they expect to be “in the next two or three years.”

Hit 100 percent is not part of the plan. Ocado expects that some elements will not deserve automation. Wine bottles and watermelon are very heavy so that the current suction cup can be treated, and the clutch may cause damage. OCADO is developing a supplement for wine bottles, because it treats many of them, but it leaves watermelon for humans – developing a tool dedicated only to one element that is not worth it.

But things can change. when freedom OCADO visited us throughout those years, we saw an early initial model for OGRP, long before it was ready for operation. “Nothing prepares a robot just like a bag of orange,” we wrote at the time, with a highlight of the boundaries of technology: the bags move unexpectedly, there is no easy point to seize a suction cup, and leave you very strong fist with juice and not fruits. Okado agreed that this was far from their reach, but after eight years, Matthews told me that the robots had discovered this themselves. Artificial intelligence models that support their programs (which Matthews call “the cousins” of the models of artificial intelligence on the headlines of the newspaper elsewhere) have not been trained on bags of fruit, but after they learned the experience that they can attach the suction cup to the right point on the poster and raise the entire bag from there, and cancel the storage of a new skill for each robot through the range.

OCADO is looking for automation opportunities almost everywhere in the warehouse. There are still workers separating the shipments of products and loading them in the boxes on the network, but I quickly told the company that the company is working on new automation for that. Others download heavy metal attachments to the outlets issued, but there is a mobile robot in the development of this function as well. It may be the safest of the actual motivations – while OCADO is investing in both Wayve and Oxa, emerging companies in the UK working on independent leadership, Matthews does not see delivery processes becomes fully mechanism any time soon. This, after all, is the only point in the process facing customers, and Matthews does not seem very attractive to a future as customers are assigned to unload delivery trucks for themselves.

Aside, OCADO is perfectly suitable for automation because its pursuit of efficiency has already made a lot of its functions, mechanical and repeated anyway. The more efficient and focused workers, the easier it is to design a robot to take over. In addition, some of the functions that are replaced are stressful and difficult for employees at the best times, such as employees assigned to fill ice cream and other frozen foods. “You cannot literally find people who want to enter and work in the freezer,” as Matthews claims, which makes these jobs normal to automate.

But OCADO also has a certain distance from his replacement employees. It sells technology within CFCs to customers, but does not manage the same daily processes. Please visit Kroger CFC in the United States, and it may be packed on the rafters with OCADO robots, but all employees will be paid by Kroger, not OCADO – and when layoffs come, they do not manage them OCADO. Ocado itself does not cut jobs. In fact, it grows, as Matthews told me, open more sites, expand the scope of research and development, and employ more support workers.

It is clear that the additional automation, and a fewer human function, is in the future to support ocado -backed groceries. But how exactly this future will look like? Eight years ago, the automatic weapons were the promise of tomorrow, so what is the equivalent now? “Efficiency” may be one answer. OCADO works lighter, cheaper and more efficient energy than their robots, including new 3D decorative models that weigh a third of the original copies. This has harmful effects In collisionSo OCADO can reduce the size of the crash around the network, which makes it more tight and more unit, and the easiest to reduce smaller locations.

But if you ask Matthews, it will be difficult to predict the greatest changes. The weapons only heavily from the development project to the operating infrastructure when the artificial intelligence models inside them delivered their own leap. The most important problems are not the physical problems, but analytical problems, designing smart machines enough to work through edge cases, to adapt to problems such as boxes that do not fit their rack, which causes jam. “It is not useful to solve something 90 percent of the time,” Matthews says. “Because if you have 10 percent of the time, you must pay an expensive engineer to go and cancel it, it is better to do it manually.”

These 10 percent problems are the place where OCADO is still and its customers feel the need to keep the participating people – but not from the gap that must be closed to machines.