Amazon
OEM : Retail
Amazon is guided by four principles: customer obsession rather than competitor focus, passion for invention, commitment to operational excellence, and long-term thinking. Amazon strives to be Earth’s most customer-centric company, Earth’s best employer, and Earth’s safest place to work. Customer reviews, 1-Click shopping, personalized recommendations, Prime, Fulfillment by Amazon, AWS, Kindle Direct Publishing, Kindle, Career Choice, Fire tablets, Fire TV, Amazon Echo, Alexa, Just Walk Out technology, Amazon Studios, and The Climate Pledge are some of the things pioneered by Amazon.
Assembly Line
Predicting congestion in fleets of robots
Many Amazon fulfillment centers use mobile robots to move shelves, retrieve products, and deliver them to workers for sorting, reducing the need for employees to walk long distances. For simplicity and scalability, the path-planning algorithm those robots currently use focuses on individual agents and ignores interactions between multiple agents.
In a paper we presented at this year’s International Conference on Robotics and Automation (ICRA), we propose a deep-learning model that can predict congestion on the floor in real time. We tested the model’s predictions in simulations of two downstream applications: dynamic path planning in sortation centers, where our model improved throughout by 4.4%, and travel time estimation, where it improved the mean absolute percentage error by 30% to 40% relative to the current production methods.
CoreTigo Receives Investment from Amazon Industrial Innovation Fund to Accelerate Industrial Connectivity
CoreTigo, an industrial wireless automation solution provider, announced today that Amazon has joined their Series B1 round of fundraising. Amazon’s Industrial Innovation Fund investment brings CoreTigo’s Series B1 round to a total of $18M and joins CoreTigo’s strategic investors, including, Emerson Ventures, Qualcomm Ventures LLC, and Verizon Ventures, among others.
🦾 Amazon’s New Robots Are Rolling Out an Automation Revolution
Proteus is part of an army of smarter robots currently rolling into Amazon’s already heavily automated fulfillment centers. Some of these machines, such as Proteus, will work among humans. And many of them take on tasks previously done by people. A robot called Sparrow, introduced in November 2022, can pick individual products from storage cubbies and place them into larger plastic bins—a step towards human-like dexterity, a holy grail of robotics and a bottleneck in the automation of a lot of manual work. Amazon also last year invested in a startup that makes humanoid robots capable of carrying boxes around.
Amazon’s latest robots could bring about a company-wide—and industry-wide—shift in the balance between automation and people. When Amazon first rolled out large numbers of robots, after acquiring startup Kiva Systems and its shelf-carrying robots in 2012, the company redesigned its fulfillment centers and distribution network, speeding up deliveries and capturing even more business. The ecommerce firm may now be on the cusp of a similar shift, with the new robots already starting to reshape fulfillment centers and how its employees work. Certain jobs will be eliminated while new ones will emerge—just as long as its business continues growing. And competitors, as always, will be forced to adapt or perish.
Amazon Turns to AI to Weed Out Damaged Goods
The AI checks items during the picking and packing process. Goods are picked for individual orders and placed into bins that move through an imaging station, where they are checked to confirm the right products have been selected. That imaging station will now also evaluate whether any items are damaged. If something is broken, the bin will move to a worker who will take a closer look. If everything looks fine, the order will be moved along to be packed and shipped to the customer.
Amazon so far has implemented the AI at two fulfillment centers and plans to roll out the system at 10 more sites in North America and Europe. The company has found the AI is three times as effective at identifying damage as a warehouse worker, said Christoph Schwerdtfeger, a software development manager at Amazon.
🦾 Amazon releases largest dataset for training 'pick and place' robots
In an effort to improve the performance of robots that pick, sort, and pack products in warehouses, Amazon has publicly released the largest dataset of images captured in an industrial product-sorting setting. Where the largest previous dataset of industrial images featured on the order of 100 objects, the Amazon dataset, called ARMBench, features more than 190,000 objects. As such, it could be used to train “pick and place” robots that are better able to generalize to new products and contexts.
The scenario in which the ARMBench images were collected involves a robotic arm that must retrieve a single item from a bin full of items and transfer it to a tray on a conveyor belt. The variety of objects and their configurations and interactions in the context of the robotic system make this a uniquely challenging task.
📦 How AWS used ML to help Amazon fulfillment centers reduce downtime by 70%
The retail leader has announced it uses Amazon Monitron, an end-to-end machine learning (ML) system to detect abnormal behavior in industrial machinery — that launched in December 2020 — to provide predictive maintenance. As a result, Amazon has reduced unplanned downtime at the fulfillment centers by nearly 70%, which helps deliver more customer orders on time.
Monitron receives automatic temperature and vibration measurements every hour, detecting potential failures within hours, compared with 4 weeks for the previous manual techniques. In the year and a half since the fulfillment centers began using it, they have helped avoid about 7,300 confirmed issues across 88 fulfillment center sites across the world.
Amazon’s New Robot Can Handle Most Items in the Everything Store
Sparrow could shift the balance between humans and machines in the company’s warehouses, using machine learning algorithms and a custom gripper. Sparrow is designed to pick out items piled in shelves or bins so they can be packed into orders for shipping to customers. That’s one of the most difficult tasks in warehouse robotics because there are so many different objects, each with different shapes, textures, and malleability, that can be piled up haphazardly. Sparrow takes on that challenge by using machine learning and cameras to identify objects piled in a bin and plan how to grab one using a custom gripper with several suction tubes. Amazon demonstrated Sparrow for the first time today at the company’s robotics manufacturing facility in Massachusetts.
How a universal model is helping one generation of Amazon robots train the next
In short, building a dataset big enough to train a demanding machine learning model requires time and resources, with no guarantee that the novel robotic process you are working toward will prove successful. This became a recurring issue for Amazon Robotics AI. So this year, work began in earnest to address the data scarcity problem. The solution: a “universal model” able to generalize to virtually any package segmentation task.
To develop the model, Meeker and her colleagues first used publicly available datasets to give their model basic classification skills — being able to distinguish boxes or packages from other things, for example. Next, they honed the model, teaching it to distinguish between many types of packaging in warehouse settings — from plastic bags to padded mailers to cardboard boxes of varying appearance — using a trove of training data compiled by the Robin program and half a dozen other Amazon teams over the last few years. This dataset comprised almost half a million annotated images.
The universal model now includes images of unpackaged items, too, allowing it to perform segmentation across a greater diversity of warehouse processes. Initiatives such as multimodal identification, which aims to visually identify items without needing to see a barcode, and the automated damage detection program are accruing product-specific data that could be fed into the universal model, as well as images taken on the fulfillment center floor by the autonomous robots that carry crates of products.
Using graph neural networks to recommend related products
In experiments, we found that our approach outperformed state-of-the-art baselines by 30% to 160%, as measured by HitRate and mean reciprocal rank, both of which compare model predictions to actual customer co-purchases. We have begun to deploy this model in production.
The main difficulty with using graph neural networks (GNNs) to do related-product recommendation is that the relationships between products are asymmetric. It makes perfect sense to recommend a phone case to someone who’s buying a new phone but less sense to recommend a phone to someone who’s buying a case. We solve this problem by producing two embeddings of every graph node: one that characterizes its role as the source of a related-product recommendation and one that characterizes its role as the target. We also present a new loss function that encourages related-product recommendation (RPR) models to select products along outbound graph edges and discourages them from recommending products along inbound edges.
Electra raises $85M to electrify and decarbonize iron and steelmaking with no green premium
Electra, a green iron company, has raised $85 million to produce Low-Temperature Iron (LTI) from commercial and low-grade ores using zero-carbon intermittent electricity. Electra’s process emits zero carbon dioxide emissions and carries zero green premium, meaning it will cost the same or less than existing production methods powered by fossil fuels.
Electra, founded by entrepreneurs with decades of experience developing complex electrochemical systems, has created a novel process to electrochemically refine iron ore into pure iron at 60 degrees Celsius (140 degrees Fahrenheit) using renewable electricity and then convert the iron to steel using the existing infrastructure of electricity-powered arc furnaces. By comparison, 69% of steel today is made at approximately 1,600 degrees Celsius (2,912 degrees Fahrenheit) using coal, emitting about two tons of carbon dioxide for every ton of steel produced.
Data-driven fault identification is key to more sustainable facilities management
HVAC units are central to a building and constitute roughly 50% of a building’s energy consumption. As a result, they are well instrumented and generally follow a rules-based approach. The downside: this approach can lead to many false alarms and building managers rely on manual inspection and occupants to communicate important faults that require attention. Building managers and engineers focus significant time and budget on HVAC systems, but nevertheless HVAC system faults still can account for 5% to 20% of energy waste.
A building’s data model, and the larger building management schema, are established when the building first opens. Alerts, alarms, and performance data are issued through the BMS and a manager will notify a building services team to take action as needed. However, as the building and infrastructure ages many alarms become endemic and are difficult to remedy. Alarm fatigue is a term often used to describe the resulting BMS operator experience.
Amazon is buying Cloostermans, a mechatronics specialist in Belgium, to ramp up its robotics operations
Amazon has made a string of startup acquisitions over the years to build out its robotics business; now, the e-commerce leviathan is taking an interesting turn in that strategy as it expands its industrial warehouse capabilities. Amazon is acquiring Cloostermans, a company out of Belgium that is a specialist in mechatronics. It’s been building technology to move and stack heavy palettes and totes, and robotics used to package products for customer orders. Amazon has been using those products as a customer of Cloostermans’ since 2019 for e-commerce operations; it’s making the acquisition to ramp up its R&D and deployment in that area.
Amazon’s Janus framework lifts continual learning to the next level
“The problem with machine learning is that models must adapt to continually changing data conditions,” says Cassie Meeker, an Amazon Robotics applied scientist who is an expert user of Amazon’s continuous learning system. “Amazon is a global company — the types of packages we ship and the distribution of these packages changes frequently. Our models need to adapt to these changes while maintaining high performance. To do this, we require continual learning.” To get there, Meeker’s team created a new learning system—a framework powerful and smart enough to adapt to distribution shifts in real time.
How Amazon learned to cut its cardboard waste
David Gasperino, an Amazon principal research scientist, led the technical development of PackOpt, which is helping Amazon’s stakeholders to not only minimize the amount of “air” shipped to customers, but also helping Amazon deliver on its Climate Pledge commitment to reaching net-zero carbon emissions across its business by 2040.
“To create an optimal set of boxes, you need to select a small subset of columns to pack all of the shipments, and those columns must lead to the smallest overall box volume when you sum it all up,” explains Gasperino. It is a hard challenge — literally. “This problem belongs to a theoretical class of problems called ‘NP hard’
Amazon Shows Off Impressive New Warehouse Robots
Proteus is our first fully autonomous mobile robot. Historically, it’s been difficult to safely incorporate robotics in the same physical space as people. We believe Proteus will change that while remaining smart, safe, and collaborative.
Proteus autonomously moves through our facilities using advanced safety, perception, and navigation technology developed by Amazon. The robot was built to be automatically directed to perform its work and move around employees—meaning it has no need to be confined to restricted areas. It can operate in a manner that augments simple, safe interaction between technology and people—opening up a broader range of possible uses to help our employees—such as the lifting and movement of GoCarts, the nonautomated, wheeled transports used to move packages through our facilities.
How Amazon robots navigate congestion
“When we first started looking at it, we thought it would take more than 8,000 robots to keep an Amazon fulfillment center running,” Durham said. “There just was not enough room for them all. That’s when we said, ‘Wow, we really have to solve the congestion problem.’ And we have addressed it — we’ve gotten dramatically more efficient.”
While good work allocation and route decisions smooth traffic flow and reduce unnecessary trips, managing the actual movement of robots is also important. To simplify the task, Amazon’s cloud computing service creates the virtual equivalent of a map of a city grid, on which robots can travel ‘north-south’ or ‘east-west’. Once a robot picks up a pod, the computing service creates a route to its final destination.
Amazon launches $1 billion fund to invest in warehouse technologies
Amazon on Thursday launched a $1 billion fund to invest in companies developing supply chain, logistics and fulfillment technologies. The first round of investments will focus on technologies that can speed up deliveries and improve the safety of workers in its warehouses. Start-ups backed by the new fund include Modjoul, a company developing wearable safety technology that issues alerts and recommendations aimed at reducing injuries.
At Amazon Robotics, simulation gains traction
“To develop complex robotic manipulation systems, we need both visual realism and accurate physics,” says Marchese. “There aren’t many simulators that can do both. Moreover, where we can, we need to preserve and exploit structure in the governing equations — this helps us analyze and control the robotic systems we build.”
Drake, an open-source toolbox for modeling and optimizing robots and their control system, brings together several desirable elements for online simulation. The first is a robust multibody dynamics engine optimized for simulating robotic devices. The second is a systems framework that lets Amazon scientists write custom models and compose these into complex systems that represent actual robots. The third is what he calls a “buffet of well-tested solvers” that resolve numerical optimizations at the core of Amazon’s models, sometimes as often as every time step of the simulation. Lastly, is its robust contact solver. It calculates the forces that occur when rigid-body items interact with one another in a simulation.
Amazon Robotics Builds Digital Twins of Warehouses with NVIDIA Omniverse and Isaac Sim
How pioneering deep learning is reducing Amazon’s packaging waste
Fortunately, machine learning approaches — particularly deep learning — thrive on big data and massive scale, and a pioneering combination of natural language processing and computer vision is enabling Amazon to hone in on using the right amount of packaging. These tools have helped Amazon drive change over the past six years, reducing per-shipment packaging weight by 36% and eliminating more than a million tons of packaging, equivalent to more than 2 billion shipping boxes.
“When the model is certain of the best package type for a given product, we allow it to auto-certify it for that pack type,” says Bales. “When the model is less certain, it flags a product and its packaging for testing by a human.” The technology is currently being applied to product lines across North America and Europe, automatically reducing waste at a growing scale.
The evolution of Amazon’s inventory planning system
Forecasting models developed by Amazon’s Supply Chain Optimization Technologies organization predict the demand for every product. Buying systems determine the right level of product to purchase from different suppliers, while large-scale placement systems determine the optimal location for products across the hundreds of facilities belonging to Amazon’s global fulfillment network.
“In 2016, Amazon’s supply chain network was designed for scenarios where inventory from any fulfillment center could be shipped to any customer to meet a two-day promise,” said Salal Humair, senior principal research scientist at Amazon who has been with the company for seven years. This design was inadequate for the new world in which Amazon was operating; one shaped by what Humair calls the “globalization-localization imperative.”
A new multi-echelon inventory system developed by SCOT (a project whose roots stretch back to 2016) is a significant break from the past. The heart of the model is a multi-product, multi-fulfillment center, capacity-constrained model for optimizing inventory levels for multiple delivery speeds, under a dynamic fulfillment policy. The framework then uses a Lagrangian-type decomposition framework to control and optimize inventory levels across Amazon’s network in near real-time.
Broadly speaking, decomposition is a mathematical technique that breaks a large, complex problem up into smaller and simpler ones. Each of these problems is then solved in parallel or sequentially. The Lagrangian method of decomposition factors complicated constraints into the solution, while providing a ‘cost’ for violating these constraints. This cost makes the problem easier to solve by providing an upper bound to the maximization problem, which is critical when planning for inventory levels at Amazon’s scale.
In Amazon’s Flagship Fulfillment Center, the Machines Run the Show
More than the physical robots, the stars of Amazon’s facilities are the algorithms—sets of computer instructions designed to solve specific problems. Software determines how many items a facility can handle, where each product is supposed to go, how many people are required for the night shift during the holiday rush, and which truck is best positioned to get a stick of deodorant to a customer on time. “We rely on the software to help us make the right decisions,” says Shobe, BFI4’s general manager.
When managers wanted to figure out how many people they needed at each station to keep up with customer orders, they once used Excel and their gut. Then, starting in about 2014, the company flew spreadsheet jockeys from warehouses around the country to Seattle and put them in a conference room with software engineers, who distilled their work and automated it. The resulting AutoFlow program was clunky at first, spitting out recommendations to put half an employee at one station and half an employee at another, recalls David Glick, a former Amazon logistics executive who supervised initial development of the software. Eventually the system learned that humans can’t be split in half.
Factory Robots! See inside Tesla, Amazon and Audi's operations (supercut)
The history of Amazon’s forecasting algorithm
Historical patterns can be leveraged to make decisions on inventory levels for products with predictable consumption patterns — think household staples like laundry detergent or trash bags. However, most products exhibit a variability in demand due to factors that are beyond Amazon’s control.
Today, Amazon’s forecasting team has drawn on advances in fields like deep learning, image recognition and natural language processing to develop a forecasting model that makes accurate decisions across diverse product categories. Arriving at this unified forecasting model hasn’t been the result of one “eureka” moment. Rather, it has been a decade-plus long journey.
Amazon’s robot arms break ground in safety, technology
Robin, one of the most complex stationary robot arm systems Amazon has ever built, brings many core technologies to new levels and acts as a glimpse into the possibilities of combining vision, package manipulation and machine learning, said Will Harris, principal product manager of the Robin program.
Those technologies can be seen when Robin goes to work. As soft mailers and boxes move down the conveyor line, Robin must break the jumble down into individual items. This is called image segmentation. People do it automatically, but for a long time, robots only saw a solid blob of pixels.
The case of the missing toilet paper: How the coronavirus exposed U.S. supply chain flaws
Before executives at consumer-goods giant Kimberly-Clark rushed to shut their offices on Friday the 13th of March, they convened for one last emergency meeting. Commuting home that final time, Arist Mastorides, president of family care for North America, stopped at his local Walmart, on the edge of Lake Winnebago in Neenah, Wis., to see the emergency firsthand. Mastorides oversees toilet paper brands like Cottonelle and Scott, but that evening he could find none of his own products. “A long gondola shelf that’s completely empty of bathroom and facial tissue, I never in my life thought I would ever see that,” he says. “That’s a very unsettling thing.”
From apple juice to antibiotics: Coronavirus epidemic could cause U.S. shortages
The toll of the ongoing coronavirus epidemic in human life is already devastating enough. But as quarantines continue in China, it looks like the global economic impact of the virus could be incredibly destructive too.
China is a manufacturing superpower, supplying both critical equipment and items of convenience. With some of the country’s citizens unable to report to work and exports curtailed, there are already shortages that have some companies worried.