RoboBusiness: General Motors Interview: Human and Robot ‘Colleagues’ in Manufacturing

Dr. Roland Menassa, presenting “Robonaut2 and Next-Generation Industrial Robots” on Wednesday, November 2, 2011 at the RoboBusiness Leadership Summit in Boston, is a research technical fellow in the Manufacturing Systems Research Lab at General Motors. A 22-year GM veteran, he now leads the development of Robonaut in collaboration with NASA, while also managing global research groups developing advanced robotics and flexible tooling concepts for vehicle assembly.

Dr. Menassa spoke with Robotics Trends writer Ellen Muraskin about robots and humans in industrial settings, as a preview to his presentation at the RoboBusiness summit.

Robotics Trends: What are the challenges of designing robots that work in close proximity with humans in manufacturing environments?

Menassa: If you look at the auto industry or any automated manufacturing system you see today, you see that traditionally, we have designed people out of that system. We think that the future is different; it’s a future in which people and machines can interact more intuitively, and where robots can take more of an assistant/coworker role. So the challenge is in bringing people closer to machinery that has, till now, been considered “unsafe.”

One approach is the design and build of a different class of machines that mimics human motion and capability. That’s the premise of what we’re doing with the Robonaut “humanoid” robot-with internal sensing, mechanisms, external design, and multiple levels of monitoring to make it as safe as possible.

Robotics Trends: How close are we to that today?

Menassa: Safety standards today are well defined for industrial robots. For exceptional cases, a Task Assessment Based Risk Analysis (TABRA) is used to assess the risk. Where close interaction between robots and humans is concerned, however, there is much work to be done to define standards. In the meantime, there are things we know. We can make a robot less powerful, so it does not exceed human strength and motion capability.

Our robot, for example, uses series elastic actuators (SEA) to isolate each joint between gearbox and motor and thus achieve passive compliance-appropriately reacting to forces exerted upon, say, an arm. We achieve fine force resolution to implement a variety of impedance controls. We also design in force torque sensing at three different layers, so that if the robot makes contact with a surface, it can interpret that contact, whether it is mere shoulder rubbing-a contact sure to happen in close proximity to human workers-or a true obstacle requiring redirection of an extremity or the whole robot itself.

We also use softer materials around the body of the robot. These are some of the measures we use to ensure redundancy in our safety systems. The challenge is to identify the minimum set of safety measures that is truly needed to enable a close collaborative mode between human and robot.

Robotics Trends: What’s wrong with letting robots work in isolation? Is that an industrial concern or a political one?

Menassa: Isolation between people and robots has worked well for the last 50 years. In a body shop, for example, where the product changes every five to six years, robots will perform the same operation over and over with great precision as a result of a very structured environment.

On the vehicle assembly line, however, the environment is dynamic, unfixtured, and semi-structured, with part changes that are very common throughout the year. Parts come in a bin; they are not fixtured. People walk around adding value to a vehicle as it proceeds down the line. Here, automation is not smart enough today to track the progress of that line, so if you interrupted that process to throw a piece of automation in, you’d have to change your material delivery system to include stop stations, additional buffers, and of course a manual station used as backup in case the automation fails. It ends up being very costly.

Welding, painting, or stamping can all be successfully automated, but when it comes to assembly, it’s very clear that any automation has to coexist in a human-centric model, in close proximity. A human is perceptive enough to know where a bolt is needed, but a human doesn’t have to take the body stress to physically torque it down. A robot can finish that job. That’s an example of a hybrid model we’re entertaining.

Robotics Trends: What do “flexible” and “reconfigurable” mean in the context of robots?

Menassa: Flexible describes an environment where a robot can work within a range of known families of parts, as in, say, three models of cars on an assembly line. Now say you add a fourth vehicle to the line-introducing new parts, new materials, welding parameters that are unknown and very different.

This goes a level higher to reconfigurability, where you must be able to change over quickly to a new product with little penalty in time, cost, or performance. The auto industry’s constant changes in models and options call for that kind of robot, one that constantly adjusts to new work content. And that makes the traditional methods by which we program robots archaic.

In such a dynamic environment, the old method of reprogramming robots, either using teach pendants and/or simulation tools, will not work. In this case we are looking for more intuitive interfaces that line workers-people who are not PhDs or engineers-can use.