We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 - 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!


During the pandemic, robotics startups, particularly those occupying the logistics and transportation markets, have attracted increasing investments from venture firms. From March 2020 to March 2021, firms poured $6.3 billion into robotics companies — up nearly 50% from the $4.3 billion they invested in the comparable 12-month period a year earlier. The surge unsurprisingly coincided with an uptick in robotics shipments, with the Association for Advancing Automation reporting that orders of industrial robots in North America surged 20% year-over-year in Q1 2021.

Companies developing robots that can grasp objects in environments like factory floors and warehouses, like Ambi Robotics, Soft Robotics, and Covariant, have received a hefty portion of the windfall. Grasping being a challenging technical problem, new challengers with new approaches have entered the field recently, including Stanford spinout Rios. Rios today announced that it raised $28 million in a series A funding round led by Main Sequence with participation from Yamaha Motor Ventures, Orbit Venture Partners, Hypertherm Ventures, Morpheus Ventures, Grit Ventures, Valley Capital Partners, and others, bringing the startup’s total raised to more than $33 million.

Solving grasping

Founded in 2018 by former Xerox PARC engineers CEO Bernard Casse, Christopher Lalau-Keraly, Christopher Paulson, Clinton Smith, and Matthew Shaffer, San Francisco, California-based Rios develops and deploys robots to factory assembly lines and warehouses to automate supply chain and logistics operations. As a business, the company provides what it calls “factory automation-as-a-service” for brands across manufacturing, food services, agriculture, and biochemical, offering a subscription-based product to mechanize individual manufacturing lines.

“We have real deployments in the manufacturing, consumer packaged goods, and food and beverage sectors, and we’re generating revenues. We’ve tripled our contracted annual recurring revenues yearly [and we’ve] signed agreements with over a dozen customers,” Casse told VentureBeat via email. “The new capital will be used to deploy our robotic fleet at scale to both new and existing customers.”

Event

Transform 2022

Join us at the leading event on applied AI for enterprise business and technology decision makers in-person July 19 and virtually from July 20-28.

Register Here

Rios’ robotics stack combines an AI orchestration system (deployed in the cloud or on-premises), embedded processors, computer vision, touch sensors, and accessories like gripper arms and pincers. The company’s machines contain thousands of touch sensors to relay data that the AI orchestrator uses, along with other example data, to learn how to grasp objects it hasn’t seen before. Rios claims that its robots can detect when an object is slipping and know when, for example, a gripper arm is closing or moving. Using computer vision, they can ostensibly also learn the appearance of new objects, leveraging this capability and the touch data to adapt to new environments and respond to “dynamic events.”

On the backend, Rios employs a software-based physics engine to simulate variables like conveyor belt speed, gravity, and the torque of a robotic arm. The company says it’s able to predict to what extent a robot will be able to perform a given task, providing a way to test and validate models and AI algorithms.

“We’re providing [customers] with real-time production and inventory data to manage production lines, in-line quality testing, predictive analytics for maintenance, and real-time adaption to shifting customer demands. We collect massive amounts of data and we have our own proprietary datasets,” Casse continued.

“We’ve built a sophisticated infrastructure under the hood that allows our robots to perform a diversity of tasks and perform increasingly complex manipulation tasks. Our robots continuously learn on the job, construct models of the world, and extend or adapt these models to perform other tasks,” Rios writes on its website. “Our robot platform generates terabytes of data … that need to be readily accessible for developing machine learning pipelines and for parallel processing across multiple cloud servers. We’ve developed a proprietary massive database processing platform for accessing and manipulating data at large scale, building engineering pipelines, standardizing machine learning models, and more.”

Challenging problems

Looking to the future, 40-employee Rios says that it’s transitioning to an infrastructure in which its robots can learn to grasp and manipulate objects on their own without any human intervention. Guided by feedback through vision and touch data, the goal will be to create models of grasping and to extend these models to different objects.

But grasping remains a difficult challenge in robotics. As a recent study points out, grasping objects in a clutter is especially difficult because the target object needs to be held in a certain, often nuanced way and orientation. Prototype household robots like PR2, developed by researchers at the University of Bremen, struggle to reliably grasp objects like spoons and cartons of milk.

“[S]everal [grasping tasks] remain challenging because either the tolerance is tight or the behaviors are difficult to model,” wrote the coauthors of a 2021 survey of research progress in robotic grasping. “For instance, peg-in-hole tasks and tightening or loosening bolts require purposeful motions. Tearing up a paper towel and pouring water require [a] robot to predict the kitchen roll motion and water flowing speed in response to a motion. Opening a bottle with a locking safety cap is a task that requests the robot to predict a pressing outcome.”

Casse argues that settings like factories and warehouses provide more structure and predictability, however, lowering some barriers to grasping that robots in other domains face.

“Customers are in pain and forced to adopt automation. It’s no longer a ‘nice to have,’ but a ‘need to have.’ Today, there’s a steep surge in demand for goods, and almost anything manufactured is in short supply,” he added. “With millions of jobs unfilled, manufacturers are unable to keep up with skyrocketing consumer demand — they are in extreme pain and need a more scalable and reliable workforce. Our company’s solution means that we’re providing a robotic workforce to solve their labor shortage and labor turnover problems that the pandemic turned into a full-blown crisis.”

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.

Author
Topics