NextGen Supply Chain Interview: Crystal Parrott, Dematic

This month we spoke with Crystal Parrott, vice president of Dematic's Robotics Center of Excellence. With more than 28 years of robotics and automation experience, she leads all of the company's robotic initiatives. Parrott was previously responsible for the development of advanced robotics solutions at Southwest Research Institute.

NextGen Supply Chain: Robots are not new to industry. We're all familiar with welding, painting and material handling robots, especially in automotive. And in distribution, robotic palletizers are well established too. But it seems we are entering a new era with picking robots, which have seen very limited use so far. What's going on here?

Parrott: The short answer is we are moving from highly structured environments to very unstructured ones. Let me explain.

The types of robots you described as established perform the same operation almost every time on components that are fixtured or constrained. Software allows new patterns of painting, welding and palletizing. However, the new patterns occur in the same envelope and are not significantly different from the previous ones. Furthermore, the motions are generally repeated again and again and again.

Most picking robots today are using technologies to address the random positions of products. They may pick and place a range of individual items but the items all fit a general profile that is easy to identify and grasp; primarily boxes, mailers, pill bottles and other common configurations.

The future of picking robots is handling thousands of disparate items that come in a wide range of sizes and weights. To make this even more challenging, the robot may not know in advance what the next item is and has to adapt on the fly. Welcome to the unstructured world.

NextGen Supply Chain: So why is this shift occurring?

Parrott: We have several forces at work all at once.

On the demand side, e-commerce is changing distribution centers and the products handled. The SKU proliferation in e-commerce is on an unprecedented scale. The numbers are truly staggering. And so is the difference in shape, size, weight and other characteristics of the SKUs.

On the supply side, two quite different forces are converging.

One is the availability of labor to work in DCs. The pool, which has to accommodate turnover, is not nearly big enough. So how is the work to get done? That's where robots and especially picking robots usually come into the conversation.
The other force is the technology that goes into picking robots. The challenge is to make it possible for robots to perform actions differently than they could before and to do it cost effectively. And do it all at high speed with even higher pick accuracy.

NextGen Supply Chain: Tell us a little more about the shifts in SKUs and why that's important for robotics.

Parrott: Let's focus on the differences in the SKUs. I already mentioned pill type bottles. Most are fairly standard in shape with size variances. But now think about bottles for shampoo, vinegar or Windex. Those are three very different kinds of bottles. And that requires the robot to distinguish between them and handle each differently based on what the robot is seeing at the time. And, oh yes, the robot has to make those decisions at least as quickly as a human can. That is still no small undertaking.

NextGen Supply Chain: That said, tell us about the shifts in technology that are making that possible.

Parrott: Before we get too far down this path, I have to say that the technology is rapidly advancing and we are in the early stages of harnessing all the capabilities that it will enable. However, the current state of development is at a point that it is now possible to implement into innovative solutions that can impact operations now.

On the hardware side, there are, of course, the robot arms and the end effectors. While collaborative robots have been added in the market allowing operation alongside humans, the robotic arms are changing primarily in the how they execute the motions they are being told to make. More on that when we get to software.

The real advances here are coming in the end effectors that the robot uses to pick up and move an item. Some are soft, tactile grippers, others are fingers still others are vacuum cups, to name a few. There are many different methods of gripping and research and development is continuing in the market. The challenge is to match an end effector to the family of parts being picked. And with such a range of SKUs, that's complicated.

Cameras/vision systems are the bridge between hardware and software. These are the robots' eyes. Advanced 3-D sensors with stereo vision can provide greater detail about the items than older 2-D cameras and sensors. The improved visualization provides essential data to the vision software and machine learning or artificial intelligence algorithms to make the pick quickly and accurately.

That information translates into directions to the arm and gripper on how to move and pick the piece. Machine learning and AI are increasingly important here. There are just too many SKUs out there for humans to be able to program the required motions for each one. Using these methods, the software learns from handling parts and teaches the robot how to handle them. The intelligent solution is what identifies the difference between bottles for pills and those for shampoo, vinegar and Windex.

NextGen Supply Chain: Any advice for people interested in the new wave of picking robots.

Parrott: Oh, yes. Picking robots are available today and improving constantly. That means you should evaluate your operation and identify high volume items that lend themselves to the current robot solutions and implement it. From that learn, expand and upgrade as the technology moves forward. Always be on the lookout for the next generation. It is coming. And probably sooner than you expect.

Gary Forger is the special projects editor for Supply Chain Management Review. He can be reached at [email protected].

SC
MR