Insider’s Guide to Emerging Technologies for the Distribution Centre


What does the distribution centre of the future look like? It’s likely to include more than a few robots and autonomous vehicles with machine learning algorithms helping to make them better at decision making. But humans won’t be entirely out of the picture. There will be a fair number of them working alongside, managing and maintaining the machines that do the heavy lifting. Here are just a few of the technologies that are coming to the DC in the next 3-5 years.

Kiva-style Goods-to-Person Bots
These bots are more like conveyor replacements. They move products and materials to the workers, reducing the amount of costly and time-consuming travel that is the bane of today’s DC worker productivity.

Perhaps one limitation with this technology is that the picking operation must be designed for goods-to-person picking, which can make it more difficult to scale up and down easily.

Several companies, like Locus Robotics and 6 River Systems, are tackling picking from a slightly different angle. These bots receive an order picking assignment and then move to the product location where a worker meets the bot to pick items into the tote. This technology allows for routing orders through multiple pick zones without needing conveyor. The ability to navigate is built into the bots themselves, so they can see objects in their path and go around them or stop and wait for the object to pass.

Effidence makes a picking cart that follows a picker through the warehouse so the picker doesn’t have to push/pull a heavy load. And Bionic Hive’s Squid works in a conventional warehouse with standard pallet rack. Units climb the vertical uprights in the racks to pick cases and deliver them to picking/packing stations. 

Self-driving Vehicles and Lift Trucks
Self-driving vehicles and lift trucks have been around for years. The older, familiar AGVs have simplistic routing and low level decision-making capabilities, require fixed paths consisting of permanently mounted beacons, barcodes or magnetic tape, which are inflexible and costly to change.

Today’s AGVs are vision-guided with improved sensors and offer more autonomy. There are no pre-programmed routes or wires in the floor because the bots can see objects in their path and learn new ways around the facility through machine learning.

Co-Bots (Collaborative Robots)
Co-bots include self-driving shuttles designed for heavier payloads, enabling batch picking with help of a human with the addition of multiple totes and pick carts. These can be paired with a pick cart or mobile put wall for more efficient batch or cluster picking solutions. The software works with a number of different warehouse execution systems to enable greater autonomy and expand the type of tasks you can assign to the bots.

Lights-Out Picking Bots
Each picking is one of the hardest problems to solve inside the DC. But new solutions are getting closer to achieving what once seemed impossible. Lights-out autonomous bots often include a picking extension or arm, with varying levels of machine learning so that the bots can execute tasks and make decisions on their own.

At this level, the bots have potential applications that go beyond picking to include packing and sorter induction.

The technology that makes lights-out picking bots possible are the unique grippers and arms available. Products vary—they are sometimes soft, fragile or irregularly shaped. They can be transparent, reflective, or geometrically inconsistent.

All of these factors amount to infinite variability and a significant degree of chaos for a machine to interpret. The annual Amazon Each Picking Challenge has really accelerated learning by bringing together some of the best minds in robotics to develop solutions that may someday offer a practical and affordable way to accomplish the task.

Robotic Depalletizers
Palletizers have been around for a while, but the ability to address the more complex task of depalletizing multi-SKU and random pallets has just recently been made possible through machine learning. Kinema offers a self-training, self-calibrating software solution for robotic depalletizing. The bots 3-D sensors “look” at a pallet to determine the shape, size and weight of items. And its algorithms determine fastest, most efficient way to depalletize the items.

Augmented Reality
Second generation Smart Glasses are addressing some of the challenges (weight, battery life and overheating) of earlier models that were built more for consumer than industrial applications.

Full augmented reality (AR), where full-sized text is overlaid on top of the “real world” scene the wearer is viewing, starts to deliver on promise of this technology. Initially, the primary applications for this technology are for picking – projecting visual cues and directions for order fulfilment tasks into a wearer’s field of view.

But down the road AR could be used for tasks, such as receiving, putaway, placement of items to a put wall, training, troubleshooting and remote maintenance.

Drones in the DC
Within the DC drones can provide an effective means of inventory cycle counting, or a way to locate missing inventory in a warehouse. Wal-Mart is testing drones in warehouses and reported that its drones can check a full warehouse of inventory in about a day. That process previously took up to a full month to do manually.

An exoskeleton transfers force from the rest of the body, including chest and back, to the thighs. Third-party logistics provider GEODIS is piloting the use of an exoskeleton by Laevo. When the employee bends over, the spring pushes back so that the load on the back is reduced by 40%.

This has great implications for safety, ergonomics and productivity, but also has potential over the long-term to enable employers to tap into alternative labor pools, such as older workers and those with disabilities.

Machine Learning
Machine learning is what enables all of these technologies to make better decisions about the best route, the best way to pick up an item and the most efficient process, in order to truly optimize processes and workflows.

Much of the data we process to make decisions is still very unstructured – machines using algorithms can do a better job of making meaning from it and potentially make better decisions as a result. But what machines today lack is flexibility that is inherent in humans.

Google Brain was founded five years ago on the principle that artificial “neural networks” that acquaint themselves with the world via trial and error, as toddlers do, might in turn develop something like human flexibility.

In the not-too-distant future, your transportation fleet will make realtime decisions based on traffic, weather, expected delivery times and make constant adjustments and tweaks. And robo-execution software will oversee fulfillment decisions in your DC, optimizing workflow from end-to-end across people, processes, systems and equipment.