For years now, supply chains have sought the ability for robots to rapidly and accurately pick and place a wide range of objects. XYZ Robotics Inc. is one startup that is committed to meeting that challenge with a combination of machine learning and robotic manipulation.
Peter Yu, chief technology officer at XYZ Robotics in Allston, Mass., recently spoke with The Robot Report about how robotics challenges led to his company’s development of its XYZ Rebinning Station and Picking Station, which he claimed is faster and more accurate than the competition.
“E-commerce is labor-intensive and faces high turnover,” he said. “Our systems are autonomous, fast, and cost-effective solutions.”
XYZ, a member of MassRobotics and the Robotic Industries Association, is debuting two products in Booth 9366 at Automate 2019. The first is the XYZ Rebinning Station (XRS), an autonomous robotic turnkey system for warehouse putwall sorting and parcel sorting.
The other product is the XYZ Picking Station (XPS), a robotic piece-picking solution for warehouse goods-to-person systems with pick rate at 1,200-1,800 SKUs per hour.
XYZ Robotics built by award-winning talent
The founders of XYZ Robotics have experience in developing robotic manipulation from their academic training and various competitions.
Yu has a Ph.D. in electrical engineering and computer science from the Massachusetts Institute of Technology, and CEO and Chief Scientist Jiaji Zhou has a Ph.D. in robotics from Carnegie Mellon University.
“At MIT, I worked on physical modeling and state estimation for robotic manipulation,” Yu said.
“My work on experimental data set for pushing manipulation was a best paper finalist at IROS,” he added, referring to the 2016 International Conference on Intelligent Robots and Systems. “Our CEO also worked on pushing modeling based-on machine learning and his paper won the best paper award at ICRA [International Conference on Robotics and Automation] that year.”
“At the 2017 Amazon Robotics Challenge, I was the technical lead for the MIT-Princeton team that championed the stowing task,” Yu said. “At RoboCup in Nagoya, Japan, our system was capable of picking and placing a wide variety of novel warehouse goods and finished the task the fastest.”
In addition to connections to MIT, the company founders chose Massachusetts because of proximity to a vibrant robotics community.
“One of the advantages of being in the Boston area is that there is access to a lot of talent,” said Yu. “We founded XYZ in April 2018, and we currently have about 20 staffers here and some in Shanghai.”
The company is looking to hire top-notch employees in both robotic perception and mechanical engineering, he said.
Need for speed, versatility
While 30 seconds per pick in the Amazon Robotics Challenge was the winner, it is far slower than a rate of 3-5 seconds per pick or 720-1,800 picks per hour desired by the logistics industry, Yu noted. Thus, XYZ Robotics continues to improve its speed.
Following needs-based development, Yu and his team have kept their focus on vision-driven automation.
XYZ’s systems don’t need to be trained on new objects, which reduced the labor required to scan millions of SKUs in a warehouse. XYZ Robotics uses self-supervised machine learning to train its robots on how to grasp items inside clutters.
“We trained our vision system with data in our lab and in customers’ facilities,” Yu said. “The robot has learned a variety of items, so it is confident to pick up any [previously] unseen items.”
To physically pick up a wider range size and weights of items, XYZ Robotics has developed a tool changer for rapidly and smoothly switching between tools on the fly — within 0.7 second. Yu said his company has patent applications pending on the automatic tool changer.
In addition, XYZ’s systems can pick up objects that are transparent, have a variety of surface textures, and are in random positions in a bin or tote. Thanks to high-level planning, they can automatically reposition items for better grasping and save space in stacking items via strategic reasoning about placing.
The company’s hand-eye coordination software is designed to be easy to integrate, and it adapts to customers’ SKU distribution so that performance improves continuously, said Yu. In addition, it has a custom end-of-arm tool (EOAT) design to further increase the speed, reliability, and most importantly, the range of items that can be handled.
Need for reliability
“Our system can be in continuous operation to keep up with the e-commerce fulfillment,” Yu said. “We have recently finished several tests in the field to verify our solutions, and we will continue the work to ensure our technology and services meet customers’ needs.”
XYZ Robotics’ target customers include third-party warehouse owners and logistics providers in the U.S. and East Asia. It offers standalone vision systems for integrators in addition to turnkey solutions for end users.
XYZ Robotics is testing at customer sites in Shanghai and is looking for U.S. customers and investors for a Series A round.
“The apparel and cosmetics markets are large,” Yu said. “We want to push the boundary in terms of speed, the variety of items, robustness, and ease of use.”