January 19, 2017
Yaro Tenzer of RightHand Robotics explains how the future of robotics is arriving at the warehouse, one product at a time.
Breathless news reports about e-commerce companies such as Amazon using robots in their warehouses may make it seem like the world is quickly moving toward a future in which some business operations are entirely automated. But the reality is that humans still play a major role in picking and packing products in almost every warehouse environment. That may soon change if Yaro Tenzer has his way. Tenzer, one of the founders of RightHand Robotics, is developing an intelligent robotic system to address one of the biggest gaps in automated fulfillment today: the labor-intensive picking and packaging of single items for shipment.
PwC: Why focus on single-item fulfillment?
Yaro Tenzer: E-commerce is growing at an unprecedented rate. What most people don’t know is that this growth is putting huge strains on the warehouses that fulfill the orders. Instead of handling items one pallet or one case at a time (to deliver products to brick-and-mortar stores, for example), warehouses now must handle each item individually. This is significantly more labor-intensive—typically amounting to more than 50 percent of warehouse operating costs.
Compounding the problem, handling individual items is a repetitive task, and businesses keep telling us they can’t find the necessary workers to meet their needs. Why? Because warehouses are generally located in rural parts of the country where real estate is inexpensive, yet the surrounding population is sparse. Therefore, warehouses are typically located in similar regions and must compete for the same small labor pool. This lack of labor becomes especially problematic for e-commerce companies during peak times, such as the holidays.
This was the big insight for us: Using robots to automate the task of picking items would solve a huge problem for a large and rapidly growing industry.
PwC: How will you address this opportunity?
Yaro Tenzer: Imagine you run a warehouse, and a customer orders something online. The item is stored somewhere in the warehouse, along with many other items, perhaps in a case or a bin on a shelf.
One way to fulfill the order is to have a picker walk to the storage location and then pick the item. Another approach is to use conveyor belts to bring the goods to the picker, who then picks the specified item and packs it for shipment. Until our robots came along, there was no automated system that could reliably and rapidly pick a broad range of items. Moreover, our robots can pick not only in unlikely laboratory settings where items are perfectly ordered on an assembly line, but also in real-world factory bins where items are typically jumbled in boxes or shelves somewhat chaotically.
The technology that our team has developed enables robots to do the pick—that is, to select one item from many.
PwC: How do you enable a robot to work with a wider class of objects?
Yaro Tenzer: Our approach is in leveraging the state of the art in computer vision, and combining it with smart gripping hardware and big data. Our team has tightly integrated these technologies to enable robots to grasp a wide range of objects reliably.
PwC: Does the robot mimic the human hand, which is unique in its structure and degrees of freedom?
Yaro Tenzer: In some ways, the robot gripper is like a human hand. Human hands have rubberlike fingers with touch sensors, and the brain seamlessly combines the touch data with what the eyes see. Our robot grippers have fingers with smart compliant mechanisms and tactile sensors, and our computer (the brain) seamlessly combines the touch data with what our vision sensors detect. However, while we are adopting the principles behind the human hand, we do not mimic the biological aspect. To clarify, we’re interested in making the robot that has the best performance, not the one most like the human hand. Our team met during a DARPA manipulation challenge a few years back, and we won that challenge by developing grippers that combine smart compliant mechanisms and tactile sensors. Having a good gripper is only part of the solution. The critical component is knowing how to combine the gripper with high-level software and machine intelligence.
PwC: What challenges are you facing as you develop your solution?
Yaro Tenzer: Initially our biggest challenge was technical—getting the robot to work. But we’ve solved those problems, and what our system can do is amazing. Now our biggest problem is perception. When we talk to our customers, we’re fighting against years of roboticists who have overpromised and underdelivered. We must convince understandably skeptical people that our robots can grasp things reliably. They can.
PwC: Do you expect applications for this technology to go beyond the warehouse?
Yaro Tenzer: Absolutely! What’s exciting is that capabilities such as robotic grasping, awareness, and robot-to-human interfaces are finally maturing to provide substantial value. Gradually these tools will evolve into tangential markets like manufacturing, grocery, agriculture, and even household assistance. Enabling robots to manipulate individual items will be the underlying technology for many of these applications and more.
PwC: What types of objects can the robot grip today? What is the next class of objects that you might address?
Yaro Tenzer: Our robots can pick a wide range of objects already, and the challenge is to be reliable—as in, successfully picking every item we attempt to pick—at high rates to provide value across many different classes of items. For comparison, current warehouse workers can pick 250 to 600 items an hour with picking errors of about 1 in 10,000. Our robots can already provide value by handling any rigid and semi-flexible items from the size of a lip balm tube to a melon. Next, we’ll tackle items that have unconstrained forms, things like apparel that can deform under their own weight. These items are challenging to pick, because their shape might change significantly as a function of how they are being picked.
PwC: With the new capabilities you are developing, what are some examples of new applications for robots?
Yaro Tenzer: There are opportunities to automate more than just picking. There is QA [quality assurance], box filling, shipping, de-trashing, receiving, and many others. Over time, with such solutions, the industry can get to lights-out order fulfillment through partnerships and further technological developments. On the order fulfillment side, we will unveil one of our systems at the big show of the National Retail Federation in early January 2017.