A revolution in robotic automation for the piece picking warehouse is emerging – mobile picking robots! The combination of mobile robots and picking arms is potentially revolutionary. These solutions are just beginning to emerge. Rochester Drug Coop (RDC) has implemented a solution from IAM Robotics with future phases already planned.
Rochester Drug Coop’s Logistics Journey
RDC describes itself as a regional wholesale drug cooperative – a marriage of a traditional drug wholesaler and a buying cooperative owned by independent pharmacies. With over $2 billion a year in sales, the company ranks as the 7th largest full-line distributor in the US.
The company supports its operations with a 65,000 square foot distribution center at their headquarters location in Rochester New York; two years ago, they opened a second warehouse in northern New Jersey.
I talked to Gary Ritzmann, a Project Manager at RDC, about their journey to implement this cutting-edge technology. “About 3 years ago, we were getting ready to open our New Jersey warehouse, but Rochester was overloaded, overused, and flat out overwhelmed.”
The company needed better throughput, but also sought a solution to high employee turnover. “We do receiving and put-away during the day and pick at night.”
Orders for pharmaceuticals, health and beauty care products and home health care supplies taken during normal working orders are delivered the next day to 1300 community retail pharmacies, long-term care pharmacies, and home health care stores in the Northeast. Most customers have just-in-time supply chains and order every day.
The Rochester warehouse stocks 20,000 stock keeping units (SKUs). Fast moving products are picked out of flow racks on the floor level. But the bulk of goods – 16,000 SKUs – are picked from a 19,000 square foot mezzanine with about 50 aisles that is divided into twelve subzones. This raised level has regular shelving.
A typical order involves two to five SKUs. Cumulatively, they pick 15,000 lines per night. Pickers use Voice technology and pick one order to a tote but with the proviso that RX and OTC goods cannot be mixed in the same tote. Then workers put the totes on a conveyor for take away. On the mezzanine level, “our order pickers do a lot of walking,” Mr. Ritzmann said.
The normal night shift hours are 5:30 pm to 2:00 am. But with the high volume of work, employees often had to work one or two hours longer. RDC was paying good money, close to $15 per hour, but turnover on the night shift was close to 40 percent. It is never easy to fill night shift jobs, but “it has become much tougher in an economy with low unemployment.”
The story of how RDC came to use IAM Robotics involved a dash of serendipity. Tom Galluzzo is the CEO at IAM Robotics. “Tom’s father is one of our customers,” Mr. Ritzmann said. “He owns a pharmacy in Buffalo. He mentioned to our CEO what his son’s company was up to.” RDC decided to look at this interesting but unproven technology. After seeing it, they decided there was potential and became a Beta customer.
IAM Robotics got started five years ago; It is a spin off from the advanced research being done at Carnegie Mellon University’s Robotics Institute. Mr. Galluzo explained that we were getting robots to look at objects, pick them up, and move them around. The Holy Grail was to get them to move from point A to B and then do something useful. To do this, the robots needed to be able to see.” Their differentiation is in computer vision technology; having robots identify a wide range of products of different sizes and shapes is a very difficult technical hurdle to overcome.
IAM Robotics built dimensioning technology called Flash to solve this problem. In less than a second, Flash records the barcode, 3D dimensions, weight and a high-resolution image of any product placed inside. The dimensional data and images are put into a server. As new products are introduced, or product dimensions change, the database needs to be kept up to date. If dimensions change and the system in not updated, the robot won’t recognize the product. When this happens, a flag is created and RDC knows they need to rescan that product. Also, if the put-away crew makes a mistake the barcode scan can detect that; the robot can flag this error as well.
Mr. Ritzmann said that at RDC they make it easier for the robot by not slotting products that have nearly the same dimensions close to each other. “But that is good practice not just to make the job easier for the robots, but to make it easier for people as well.” At RDC products are selected by the robot based on dimensions and not the bar code data.
The database also understands where the products are slotted in the warehouse, and can direct the robot to within 6 to 10 inches of the product. Machine learning is used to continue to improve the ability of the robots to properly position the robotic arm and pick the products.
The picking is done with robotic arms from FANUC attached to a mobile base. The arms have suction cups at the end attached to a vacuum by a plastic tube. As the robot approaches the product, the vacuum is activated and the arm grabs the product and drops it into the tote. Using this technology, light, small, rigid items like boxes and bottles can be picked. But those products represent about 85 percent of RDC’s products.
Currently, the robots are engaged in picking in one of the subzones on the mezzanine level. Humans are still involved. The robot has a tote attached to it. The arm picks into the tote, the robot moves to the conveyor, and a person takes the products out of the tote on the robot and puts them in a different tote on the conveyor for take away.
The ROI of Mobile Picking Robots Will Improve
With new technologies, the experimentation and learning phases clearly take longer than for more mature products. The ROI from the solution will improve as robots are used in more zones, as the robot is programmed to put full totes on a conveyor lane and then to carry an empty tote, and as the robots are used in more shifts in a day. “Right now the robot is only working 3 hours per night. The rest of the time is just sitting there. So, as we expand the number of zones a robot picks in, we will see greater benefits,” Mr. Ritzmann pointed out.
Mr. Ritzmann and I also discussed the picking speed of robots in comparison to people. In their bays, the robot picks on one side of an aisle and then spins to pick an item on the facing aisle. The robot can’t do this as quickly as a person. If the layout of the facility allowed for picking on just one side of an aisle, the bot would be just as fast as a person. But the picking speeds of robots are not all that important as long as the robots are priced economically enough. You just throw more robots at the job and still get your ROI.
Mr. Ritzmann said that “Tom and his crew” have been easy to work with. Mr. Ritzmann has been through other implementations of warehouse hardware and software that have been far more difficult. “We turned it on that first night and it worked.” In particular, integration between various systems has been difficult for other technologies, but much less so for the IAM Robotics solution.
While at RDC, this is still an ongoing journey to fully leverage the capabilities of this technology, the potential is clearly huge. Amazon paid $775 million for Kiva, a provider of autonomous mobile robots, but these robots are not capable of picking. In contrast, the bots from IAM Robotics do the full job of a picker. One can only ask, what might this company be worth as the technology matures?