Robotics: On a case-by-case basis

For several years now, industrial robot suppliers have worked to penetrate warehousing and distribution applications like palletizing, picking and case handling. Early missteps resulted from poor pairings, where a robot proven in a production environment struggled with order fulfillment, or specialized control software synced up awkwardly with existing systems, or robots were introduced as a cure-all without appreciation for their impact on upstream and downstream processes. In other cases, the justification for investment didn’t hold up over time, or simply wasn’t there to begin with.

Robotic solutions and related methodologies have almost completely overcome an impressive number of these challenges in just the past two years. But rapid shifts in market pressures and customer demands have been equally impressive, and as soon as a solution and application approach harmony, the target moves again. The limited availability of labor favors robotics alternatives, but SKU proliferation remains a challenge. Nimble grippers allow robots to handle a case securely, but that SKU’s packaging characteristics could change suddenly and without warning.

This volatile environment, however, has accelerated development rather than hindered it. From their roots in rigid repetition, robots are being forced to mature into adaptable, self-teaching, highly dexterous systems that fit within warehousing and distribution environments more like intelligent coworkers than pieces of equipment. Although it has proven a quick study, the warehouse robot still has plenty to learn.

“I’ve watched the evolution of welding robots from the early stages to where they are now, and I’ve seen plenty of ‘revolutionary’ robots get bulldozed in the early days,” says Bob Hoffman, director of sales support and strategy for Grenzebach Logistics. “But now, no one would even think of installing a manual weld station. That’s where warehouse robotics is going to be one day. Once we fight through the growing pains, some day someone will say: ‘Why did we ever manually case pick?’”

Making the case
The answer is that, once upon a time, throwing labor at the problem was as good a way as any to get the job done. The scale starts tipping when speed becomes the central objective. The rate at which a person or robot can stack a case is one thing, but more important are the pace at which product from suppliers moves through a facility, the ability to quickly get staff onboard to keep pace with growth and seasonal peaks, and, ultimately, the speed with which deliveries reach customers.

“DCs are therefore getting closer to urban centers and that creates pressure to be more efficient in less space,” says Patrick Pepin, business development manager for Axium, which specializes in robotic materials handling and assembly solutions. “As you get closer to urban areas, it’s also more of a challenge to find a workforce willing to handle boxes all day, especially across three shifts.”

Exponential growth in the number of SKUs further complicates the issue. Hoffman offers the example of a food and beverage distributor that anticipates the addition of 1,000 SKUs in the coming year. “The area they serve still drinks the same amount of product, but the variety of SKUs is changing,” he says. “That means the number of SKUs per pallet keeps growing due to the increase in slow-movers. To run from pick face to pick face hand-stacking a case here and there will quickly become unsustainable.”

These challenges impact large and small operations alike, according to Ross Halket, executive director of ASD sales for Schaefer Systems International. “Little guys are struggling against the big guys who have the economies of scale, and the only way they can win is to take incremental bites out of their supply chain,” he says. “As a result, some of the smaller companies are actually spending more capital on automation than the big ones.”

Still, uncertainty can be a big hurdle for those considering robotics, and Halket says a system’s stability means not only handling all the volume, but all the variations. “If today a system works well with 1,000 SKUs, the customer needs to know they can get to 1,500 SKUs—without knowing what the new 500 will be—and be sure the machine can handle it.”

Get a grip
Perhaps more than any other element of robotic development in warehousing and distribution, the end-of-arm tooling, or gripper, is the place that can make or break a robot’s success. A robot might have advanced vision, speedy cycles, extreme precision, and smart software, but if it struggles to grasp even a small percentage of SKUs, the application is unlikely to succeed.

“It’s easy to achieve speed with robots in a production environment,” Pepin says. “In an unstructured environment you need a system efficient enough and flexible enough to handle shrink wrap, cases, cans, trays, you name it. Handling one type of packaging quickly means nothing if you can’t handle another type of packaging equally fast.”

Borrowing from lessons in manufacturing, suppliers considered automatically swapping grippers as needed, but the several seconds required to make the change were seconds where productivity was zero. If the application calls for swapping, a better practice is to install more than one robot, each specializing in a certain range of SKUs. This also creates redundancy in the event one robot is down or does need a gripper change.

In just the past couple of years, grippers have become adept at handling all sorts of cases individually, in groups or in layers. “The speed of evolution is huge,” Pepin says. “In 2009, a single-robot, mixed-case palletizing cell handled around 500 cases per hour. In six years, we can now reliably reach more than three times that.”

The demand for layer-picking solutions is particularly strong, and these applications magnify the impacts of packaging changes. Gripper technologies have had to tread carefully amidst transitions from corrugated to shrink wrap, the removal of trays, and the risk of damage to a pallet of goods from a single broken juice bottle. Tom Pollard, applications engineer for Cimcorp North America (formerly RMT Robotics), says a single gripper using vacuums, clamps and magnets can now handle items that were out of the question until very recently.

“There was no alternative; you had to go for a manual process instead,” Pollard says. “Now, a lot more situations can apply the technology. Customers wonder if the robot will still work if they take on a new product line. With a high probability we can say it will, whereas three years ago I would have had to ask more questions to qualify that.”

Grippers are so advanced that Hoffman describes them as essentially “a robot on the end of a robot.” Capable of self-programming, grippers can collect and hand off information to the robot and into the system. For a particular SKU, Pepin suggests, one supplier facility might package it in two rows of six, while a sister facility does it in four rows of three. The gripper can identify this discrepancy, react to it, and update the system accordingly.

Similarly, the gripper and robot can draw data from the system of record to adjust the vectors on how it will bring the case to the pallet. Robots building pallets should know if it’s a jar of honey that could destroy a pallet if it broke, or if it’s diapers that can be handled more rapidly, Schaefer’s Halket says. The robot can slow down or speed up on a SKU-by-SKU basis.

This approach depends on fairly robust warehouse management and control systems, but it is rare that companies have a good information technology backbone to deploy a robot, according to Alfredo Valadez, vice president of business development for Wynright’s robotics division.

Even sophisticated customers might have advanced control systems, but they are based on the rules and picking processes as they existed before automation. “When you introduce robots,” Valadez explains, “you need to add additional rules, including the details of how each case should be handled.”

There are two methodologies to address this. Valadez says it’s possible to start building a database from scratch by manually populating it, or automation can help collect SKU-specific data. At a “teach-in” station, a new SKU can be weighed and measured in terms of displacement as opposed to simple width and height, since not every item is square. Manual entry is costly, time-consuming and an opportunity to introduce errors, and automated systems can capture this information before a new or differently packaged SKU enters the system.

Goals in sight
Halket says vision is essential to creating a solution that can handle unpredictability. Lasers, sensors and cameras can combine to better detect and identify an item, which can help bypass manual intervention for simple obstacles like poor orientation, damaged labels, drops or other exceptions. Vision systems can even keep an eye on an item in motion, meaning a piece-picking robot can retrieve items in motion. For example, piece-picking robots of the past might not have boosted performance very much if a conveyed tote had to come to a stop and allow a robot to “think” about how to get an accurate grip on an item.

“Now the arm, not the base, can move along with a donor tote while confirming the product is correct and outlining the best donor piece,” Halket says. “If the piece is at an angle or there’s insufficient contrast, it can still challenge these solutions, but it’s a rapidly developing technology.”
In other applications, vision can enable mobility of the robot’s base. Daniel Theobald, co-founder and CTO for Vecna, says it starts with perception. LIDAR (light detection and ranging), ultrasound, infrared, cameras, bumpers and more help a robot perceive the world.

“Once it perceives something, the robot has to make sense of that information,” Theobald explains. “Then it can navigate and maneuver itself and is not dependent on a hard-coded path. Once you have a robot that can get from point A to point B, the next challenge is manipulation.”

Theobald describes a project where a mobile robot can autonomously pick a 50-pound box or an individual pen out of a box. It can then put the box on a pallet or the pen on a shelf.

Bends in the river
No matter the task a robot is asked to perform, its deployment is sure to impact processes upstream and downstream of that task. “Early on, when robots entered the warehouse, they were often looked at as an island where existing systems could simply feed the robot,” Valadez recalls. “Lots of early failures were due to that mindset.”

Especially if a robot can pick at three times the speed a person, it’s critical to ensure downstream functions can handle that velocity. A complete solution will include consideration of these impacts as well as the necessary control software to coordinate neighboring processes.

Cimcorp’s Pollard says robots have also improved a facility’s overall layout. Once the robot is in place, you can springboard into other improvements, he says. Upstream, suppliers can use an automated order fulfillment system as a means to schedule production, rather than on a forecast. Downstream, it used to be that because an order fulfillment system was less predictable, an operation would often stage product near the dock then load from that buffer area to the truck. With a layer-picking robot, it is possible to delay the picking until the truck comes, Pollard says, then pick the entire contents of the truck while it’s parked.

“The buffering of products at the dock is starting to go away,” Pollard says. “Efforts to establish a just-in-time order fulfillment process can limit inventory. We used to see lots of companies staging 30 or 40 trailers in the yard that won’t even leave until tomorrow. Then they have inventory in the yard they would have rather shipped to someone else, but they can’t because it’s tied up. Instead, just pick it at the last minute.”

Once a robot is installed and running, benefits continue as its owners learn and experiment with the system. “Once the customer takes ownership, you start to see the really exciting things happen,” Hoffman says. “They might try something the team never thought of when we designed the system, and they suddenly gain a jump in production just by changing one parameter. They start taking pride in finding a half a percent here and there, and times five million cases per year that adds up quickly.”

Companies mentioned in this article
Axium Solutions, axiumsolutions.com
Cimcorp Automation, cimcorp.com
Grenzebach - Logistics, grenzebach.com
SSI Schaefer Systems International, ssi-schaefer.us
Vecna Technologies, vecna.com
Wynright Corp., wynright.com