Robotics: Training robotic pickers to do their piece

Amid rapid advances and adoption of robotic technology in materials handling, it can be hard to keep up with all the tipping points. Grippers, vision systems, navigation solutions and software are combining to make once fantastical robotic applications all but inevitable. But, even as they offer solutions to age-old challenges, concepts like robotic piece picking are also bound to present sizable tests.

When it comes to piece picking, humans have the advantage of highly dexterous manipulation, versatile product identification and intuitive decision-making. Robots offer consistency and predictability, but they will need greater speed and flexibility to keep up with the inconsistent and unpredictable world of e-commerce fulfillment.

Researchers, solution providers and end-users are still in the very early stages of identifying the opportunities, developing highly customized solutions and quantifying the results. However, because the optimal piece-picking robotic system will depend on efficient and connected processes, many operations will build comfortable homes for robots long before they are ready to move in.

“In the past, there were attempts to simply replace a person with a robot, which rarely works,” says Manfred Preiss, vice president of global sales for Schaefer Systems International (SSI). “There has to be the right environment surrounding the robot, and we have many of the pieces today. It’s now a matter of putting those pieces together in the best way.”

The pick, not the litter
Preiss says the cost and performance of robots, sensors and other components continue to get better, but the state of the labor market has become an even more powerful driver for many companies considering robotic alternatives. It’s not just the cost of labor, he says, but the availability. Combined with the order profiles common to e-commerce, it’s a recipe for increased costs and errors.
“In e-commerce you might see 35% to 55% single-line orders, which is a lot of work when you do it manually,” he says. “If you want the highest quality picking with the cheapest people, it just doesn’t go together.”

Similarly, robotic piece-picking solutions are rarely a direct substitute for humans in conventional picking areas. Instead, Earl Wohlrab, palletizing and robotics systems manager for Intelligrated, says robots will make gradual gains in these environments by targeting specific elements of the picking process.

“I see robots offering value by providing mobility in the spaces between tasks,” Wohlrab says. “This will enhance the amount of tasks a picker can do within a work zone, where he or she might work alongside a mobile, collaborative robot.”

For operations that cannot justify a robotic goods-to-person solution or automated storage and retrieval system (AS/RS), this kind of mobile robot could support picking, replenishment and packing without requiring major adjustments to existing processes or infrastructure. That said, Wohlrab encourages customers to consider wider changes as they pursue robotic picking solutions.

“If you simply want to inject automation into a process, as opposed to thinking about new ways to do the same thing, you’re missing an opportunity,” Wohlrab says. “The folks who embrace things like wireless networks and connected systems will have an easier time because they will be able to more broadly integrate the operation from receiving to shipping, as opposed to those who continuously compartmentalize and only pursue solutions that make each of those processes better.”

For example, even as vision systems become more capable, robots benefit greatly from consistent labeling. Wohlrab says too many companies rely on lots of manual labor at receiving to ensure accurate labeling for automation throughout the rest of a facility. When a human takes something off the back of a truck, there’s a good chance they can tell the difference between grapes and cherries, he says. Once manually identified, items can move more seamlessly through automated systems.

“But is that the best time to apply a label, relying on a guy with a stack of boxes on the back of a truck, or would you rather label at the manufacturer when they are producing 150,000 units at a time?” Wohlrab asks. A conversation with the manufacturer might enable accurate and speedy materials handling as soon as product arrives at the facility. In the meantime, there are several touch points for a human to identify, verify and move. More often than not, Wohlrab says, those manual activities keep getting handed downstream.

“A lot of folks have been focused on shipping efficiencies, but because of the nature of receiving there are four or five times as many people there,” he says. “Shipping is a lot less stratified operationally, so moving forward, receiving is where we’ll have to be smart and sharpen our pencils. Automation has always been a desire, but it’s moving toward a necessity.”

Fixing the path
From receiving, mobile robots offer several new takes on the movement of goods to storage and pick areas, from pick areas to packing stations, and from there to shipping. While it is clear that a robot is better suited to transporting materials and a human is best kept picking as much as possible, it’s not as clear, however, precisely how those functions should be divvied up.

If eaches are manually or automatically de-trashed at receiving and placed in a bin, an automatic guided vehicle (AGV) might ferry them to a pick location, or straight to packout. A robotic picking system might both pick and pack. People and robots might occupy the same aisle, collaborate on the same order, or specialize in the handling of certain SKU or order profiles. Of all the possible paths, experts agree that there are a few configurations that seem promising.

Many people are familiar with goods-to-person systems, including high-density AS/RS and shuttle systems, autonomous carts, or conveyor and sortation systems. Preiss sees opportunities for goods-to-robot systems, which capitalize on the established performance of goods-to-person solutions and stationary robots.

Preiss describes one customer’s system of piece-picking robots stationed at a pocket sorter. Surprisingly, the robots are fed by manual processes in this case, but whatever the upstream mechanism for presenting goods to the robot, it can handle the retrieval and depositing of as much as 90% of the customer’s SKUs.

“Many customers also want to reserve the right to manually put to the pocket sorter for that remaining 10% and to consolidate multi-line orders,” Preiss says. “The customer’s previous system was designed to pick for retail, and they were adding e-commerce capabilities. They went from 1,500 orders to 15,000, but were still handling essentially the same number of pieces. They now have a 15-minute order turnaround.”

Integrating a piece-picking robot with a mobile platform introduces a number of additional considerations. Mike Oitzman, product line manager for mobile robots Adept Technology, says the questions center on the need for the mobile robotic base to support not just a robotic arm but also to buffer the orders as products are picked. This might look like a set of totes, a traditional pallet jack, a cart system, or perhaps a mother/daughter system that allows an empty set of orders to dock with an available robot – immediately after a cart with fulfilled orders decouples and transports itself to packout.

One of the key system constraints for mobile robots, Oitzman says, is the requirement for onboard processing and having enough power to last a whole shift. Even more critical is that safety concerns limit the weight a collaborative robotic arm can handle, thus limiting the size of product to be picked.

“You could make the broad assumption that people are generally not willing to retrofit existing warehouses,” Oitzman says. “Therefore you have to make sure that any mobile robotic system can safely work around existing employees.”

Taking control
Aside from nimble grippers and sophisticated vision systems, mobile robotic pickers will also demand robust information technology and control systems. Alfredo Valadez, vice president of business development for Wynright robotics division, says most current warehouse management systems (WMS) and warehouse control systems (WCS) are not intended to handle some of the specific parameters a mobile robot needs.

“For example, they don’t now have traffic control capabilities, because they don’t need them,” he says. “It will be complicated, because as those sequences are happening for each robot, you might have hundreds of those instances happening at the same time. Hardware is one thing, controlling it is an entirely different issue.”

Tom Galluzzo, founder and CEO of IAM Robotics, agrees, and says it’s difficult to get a robot to do anything more than a single-step task without exponentially increasing complexity. Fail at any one step, he says, and the whole process fails.

“Even if the robot can do each step with 99.9% accuracy, multiply all the steps together and you might end up with a 50% chance of success,” Galluzzo says. “A mobile piece-picking robot needs to be able to pick not only from a flat shelf, but from a flow rack, pallets and other configurations you find in a warehouse.”

Galluzzo says his company’s experience in mobile picking started with the idea of a robot facing the shelves of a retail environment, specifically as a stocker in a pharmacy. In this type of application, an arm picking from a static shelf can pick 100 items per hour, provided they are small, rigid items like boxes and bottles. In a warehouse, Galluzzo says slotting will be a high priority, and might create zones of SKUs he calls “robot-friendlies.”

“That is not to say the robot will work in a separate zone,” he says. “More often than not, automation and manual laborers will work together in the same area. You don’t want to take that off the table since it’s still a good way to handle peak times, with both human and robotic assets able to access the same SKUs.”

That flexibility will allow the manager at the controls to assess people, robots and assets and redeploy as needed. Galluzzo envisions a gestural interface that could allow the manager to drag and drop a robot to address expedited orders or exceptions. Robots can’t always resolve exceptions as aptly as a person, but the control and order management systems will not care about the difference between the two.

“As robots get better at handling a wider array of SKUs, rigid but not regular products will be the next horizon,” Galluzzo says. “It will be some time, and 95% of the major breakthroughs in this space will come from the software side.” 

Companies mentioned in this article
Adept Technology
IAM Robotics
Intelligrated
Schaefer Systems International (SSI)
Wynright Corp.