Where Industry Meets Innovation

  • Contact Us
  • sign in Sign In
  • Sign in with certificate
mit campus

Resources

Search News

  • View All
  • ILP News
  • MIT Research News
  • MIT Sloan Management Review
  • Technology Review
  • Startup Exchange

ILP Institute Insider

May 21, 2018

Creating a robot with a dexterous grip

Alberto Rodriguez emphasizes robots' need to perform actions that humans do without thought for more versatile, cost-efficient use.

Steve Calechman

Robots come with a certain buzz. Companies want to take advantage of them, and for good reason. They hold the promise for greater efficiency and productivity. But the benefit doesn’t come from merely turning them on and having them perform a repetitive task. That’s not enough time and money being saved, according to Alberto Rodriguez, assistant professor of mechanical engineering. The key is in having robots that can adapt to new objects, exploit tactile sensing, and use hand motions and the environment in order to grip and re-grip, actions that humans do without thought. It’s the focus of his lab, and getting that breakthrough would be “an enabling technology,” he says, allowing the same robot to do more, from picking up various boxes to restocking shelves to cleaning up at night. It ultimately means a new way of not only designing a warehouse but also allowing people to take on more cognitive challenges, creating greater value for a company and a bigger return on the investment.

The need to adjust
Picking up objects is a fundamental part of any physical task in any warehouse or store, one that is not going to be phased out. It’s also a job that a robot can already do in a repetitive fashion. The future lies in dexterity, and a robot’s ability to differentiate between things and manipulate each one accordingly. Rodriguez keys on two elements. One is the generality of grasping. There are hundreds of objects in a factory. It would be impractical to teach a robot every one of them, he says. The intention is to have the robot learn a sampling, and from that to generalize to a larger number of items.



Alberto Rodriguez
Walter Henry Gale (1929) Career
Development Assistant Professor of
Mechanical Engineering

Combined with that is understanding how to manipulate objects. When grasping, a robot must avoid colliding with the environment, but it also needs to know when and how best to leverage it. A robot can change grips through the mobility of its fingers, but it can take advantage of its surroundings as well, by fully or partially resting an object, in essence using a floor or a wall as an extra finger. It’s a move that people intuitively make, and while the capability isn’t fully there yet, Rodriguez says that it’s developing, and an efficient use of the environment increases robot dexterity without requiring expensive parts.

The second area of his research works in conjunction with mobility, tactile sensing. As soon as something is touched, information about how hard or soft to grip is transmitted to the fingers. It’s another aspect that people can take for granted, but Rodriguez says to imagine trying to work a pen or flip pages in a book while wearing gloves. A simple process becomes clunky and inefficient as, “You end up having sticks attached to your wrists,” he says.

Rodriguez is working with other MIT researchers, in which sensors, powered by cell phone cameras inside the robot’s fingertips, provide high resolution tactile feedback with minimal cost and instrumentation. The robot can see the contact points and learn how to adjust pressure and the characteristics of different materials, with the goal of being able to incorporate the sensory information in real time. It’s a promising technology, he says, and it’s “probably one of the most important challenges that this community has to solve in the next few years” if robotic manipulation is going to be successful.



The look of the future
Warehouses and stores have needs that automation can address, but the best robot is one designed based on practical industry knowledge. For the last three years, Rodriguez has participated in the Amazon Robotics Challenge and from that he’s gained such insights. Assembly lines require robots that can take just a few objects and pick up, re-grasp and place them back down millions of times. Warehouses, on the other hand, need robots that can pick up, identify, and place thousands of different objects, many of which they have never seen or handled before. “The need for generalization and adaptability is very different,” Rodriguez says, “and so, the required technologies are consequently different too.”

With the capability to pick up a variety of things, a robot would be able to sense how objects are stocked and packed. It could move them. It could realign and reorganize them. It could even clean up at night, with little to no supervision. The upshot is that it could do all of this without having to be reconfigured for each individual task, decreasing unnecessary downtime. It would also change how product manufacturing and distribution is conceived and organized. A company could design and build a space for its own rapidly changing needs without having to start from scratch with each new product. “More than that, robotic dexterity would give enough flexibility to enable factories to manufacture products that have not been designed yet,” Rodriguez says.

Invariably, any use of robots raises concerns of how they’d affect jobs. Rodriguez doesn’t dismiss those concerns, but he’s optimistic. He sees robots as an invaluable complement to a workforce, performing the manual, boring tasks, and freeing up people to do the more cognitive ones. In that way, they’re like computers, which also brought initial fear and hesitancy. But they’ve allowed people to do more and to do what at one time was deemed unimaginable. “No employee that uses a computer would want to do their jobs without one, just like no crane operator would want to work without a crane,” he says. “Computers are tools of productivity, and I see robots in the future playing a similar role.”