Future of Work Automation — 3.3 Amazon’s Warehouses Robots/ Machines

7 min readSep 9, 2019

There are 5 types of Amazon warehouse/ delivery robots I identified through my research. They perform different tasks in the warehouse and logistics business. These robots perform heavy-duty tasks well when clear instructions are articulated (e.g., lifting heavy objects or finding the best route within a pre-defined area with set rules). As technology further develops (AI and machine learning), we will see how humans and robots work together to make the logistics business more efficient.

1. Robotic arm

There are two types of robotic arms currently used in Amazon warehouses. The first type is an industrial massive lifting robot — “Robo Stow.” It is a 6-tonnes robotic arm that lifts up to 3,000 pounds of pallet from one floor to another floor. The second type is a smaller robotic arm that stacks pallets/ containers.

2. Amazon (Kiva) Robots and Grid System

Amazon acquired Kiva Systems for $775 Million in 2012. It was the second latest acquisition Amazon made at that time. In 2014, Amazon rolled out its Amazon Robots (designed based on the Kiva Systems technology) to its warehouses.

Amazon system works on a flexible design — the storage is dynamic in a vertical stack with compartments (pod).

In traditional warehouses, goods are stored in a static place and moved around using either a conveyor belt system (one-directional) that depends on a rigid workflow. In some warehouses, goods are stored then driven by human-operated machines. Think of your last visit to Ikea, they have human-operated forklifts to get pallets of compacted furniture parcels from one area to another.

Amazon system works on a flexible design — the storage is dynamic in a vertical stack with compartments (pod). Once an order is placed in the system, the system locates an Amazon Robot with directions to retrieve the item and bring it to the human packer. The robots navigate in the warehouse by following the barcode stickers on the floor. Each robot has a motion sensor. It can rotate, scan the QR code, lift the pod through a corkscrew action (up to 3,000 pounds), follows navigation directions to retrieve a pod and deliver it to a human picker. It goes to the docking system to recharge itself for 5 minutes each hour.

The robots save human pickers time to find the items in the warehouse. Instead, human pickers spend time focusing on using their judgments to see if there are defects on these items before sending them to the Packing Station.

3. Pegasus and Xanthus Robots

Pegasus

After a human worker puts a parcel on the Pegasus robot, the robot will route itself to the right hole. The robot knows where to go because the floor/ highway has a few hundreds of rectangular holes to guide it. Each hole corresponds to the zip code. Once the orange robot gets there, it will use its mini-conveyor belt to slide the parcel down a chute/ slide to the floor below. The chute connects to the right truck for delivery.

Beyond completing its mission, the Pegasus robot will line up at the periphery to wait for the human worker giving its next mission.

China’s eCommerce companies (e.g., Alibaba) have robots similar to Kiva and Pegasus. They have been operating at an enormous scale to satisfy the demand.

Xanthus

This is a brand-new robot developed by Amazon that offers “modularity.” This robot can be outfitted with different attachments for different functions. For example, it can carry a parcel. It can also stack up containers (think Amazon Robot above). If you attach a mini-conveyor belt (think Pegasus Robot above), it can do sortation.

This design is to bridge the separation of a fulfillment center and a sortation center.

You may be thinking — what is so special about a multi-functional robot. The essence of this design is to bridge the separation of a fulfillment center and a sortation center. The core function of a fulfillment center is to pick and pack the goods into a parcel. On the other hand, the primary purpose of a sortation center is to sort these assembled boxes into the right zip codes. Xanthus potentially can purpose these two functions and simplify the processes and consolidate the fulfillment and sortation center into one.

4. Packing Machine

Amazon has added a new machine that can scan and box items. This machine is called CartonWrap, which is made by Italy’s CMC. Currently, these machines are deployed in selected warehouses, but it is expected to be rolled out to more warehouses.

The machine can box 600–700 boxes per hour, which is 4x-5x the rate of a human packer. The cost of the device is estimated at $1 million. Amazon is expected to recover the upfront cost of each machine in two years. The machine can potentially replace 24 jobs in each of Amazon’s fulfillment centers. If it is rolled out to the 55 centers in the US, 1,300 jobs could be replaced.

Implication

This robot that can scan and pack is a significant milestone for Amazon. People think robots can do simple/ less sophisticated tasks only. For example, transporting goods (rectangular shaped parcels or pallets) and following a set of “programmed rules.” But this robot is different — it can do “smart” things, such as using fine motor skills to pick up irregular shaped items and box them. Warehouse automation is heating up in Amazon. In the future, machine learning can help these machines/ robots to “learn” through observations and develops a best practice in each iteration of education.

5. Delivery Robot — Amazon Scout

Amazon Scout is an autonomous electric delivery created in Amazon’s R&D lab. It is a small robot that moves slowly on a sidewalk, accompanied by an Amazon employee. This service is available in Snohomish County, Washington, only.

6. Summary — Exploring the Human and Robot Collaboration

Work Environment Safety

Amazon thinks a fully automated warehouse is at least 10 years away. For the time being, humans and robots are collaborating together in the warehouse — adapting to each other’s strengths. Work environment safety for human workers is a crucial topic. Amazon recently acquired Canvas Technology that specializes in “spatial AI.” Spatial AI allows robots to navigate safely around humans. There are a few potential use cases:

Amazon Robots can use spatial AI to learn and update a shared map using sensors and simultaneous localization and mapping software (SLAM). Robots do not need a prior map and can adapt in a dynamic environment around human

2. Amazon Scout can use Canvas Technology’s safety systems

3. Other technologies: Canvas Autonomous Cart and Vesta Home Robot Project

Job Creation and Reskilling

Amazon posted a video on robots create jobs in the warehouse with an example of a warehouse employee becoming a robot traffic control specialist. Amazon undoubtedly is trying to make a point that Robots are adding value and creating jobs. Robots are designed to perform some existing tasks. But humans have to develop/ enhance other skills to perform new tasks or enhance existing tasks. I will have another blog post on reskilling workers later.