Smarter, Smaller, Safer Robots

Robots’ speed and brawn—and the fact that they don’t get bored or suffer repetitive-stress injuries—make them a no-brainer for manufacturing. But they require a commitment. To be integrated into a factory, they need to be precisely placed (to within millimeters), bolted to the floor, and fenced off to keep workers out of harm’s way. At least, that’s what used to be required. “In manufacturing, it’s hard to move a robot, but it’s getting less hard,” says Julie Shah, an associate professor and the director of MIT’s Interactive Robotics Group. Shah is leading research into a new generation of adaptive robots that are smarter, smaller, and safer. They’re also more flexible—they can learn new tasks from workers as the need arises. They’re more efficient themselves and they make workers more efficient too. Shah’s research suggests that these new robots represent the beginning of an important shift in manufacturing, from a process in which people work near large, dangerous automated machines to one in which they work alongside—or even with—machines. Adaptive robots are already creating new efficiencies for manufacturers and changing how companies build things, including factory floors themselves. And they are proving far more collaborative than their predecessors, enhancing engagement and job satisfaction among their human colleagues. 3 Ways Adaptive Robots Improve Manufacturing Courtesy of Rethink Robotics Robots use optics to notice when supplies are low and alert human coworkers, decreasing downtime. Algorithms enable them to recognize, pick up, and manipulate items even if the items are placed imprecisely, increasing efficiency. Sensors keep their arms from hitting people who are close by, increasing safety. All this is happening mainly because the fences around robots are coming down. The robots of the past needed ample clearance—they didn’t know when people were close by, and a mechanical arm swinging at high speed to grab a part could easily maim or kill someone. So laws requiring barriers between workers and machines were implemented in countries around the world. But sensor technology similar to the collision detection systems in new cars now makes it safe for robots to move and work among people more freely. When they detect objects nearby, they respond to prevent collisions. The technology is sound enough that governments are removing their fencing mandates. Other technological advances have helped too, including better actuators and optics and more-sophisticated software. The older generation of robots needed to have parts placed in a precise manner before they could pick them up and do their work, but adaptive robots aren’t so finicky. For example, cameras and object recognition algorithms help robots see the parts they’re supposed to handle, regardless of how those parts are placed. Advanced arms and hands can twist and turn to grasp objects and manipulate them into the proper position. Robots can also recognize when trays are empty and alert workers to refill them, waiting patiently until someone does so. All this reduces bottlenecks on the assembly line and saves precious time. What’s more, even as robots are getting smarter, their cost is dropping. The Question Is Always: What Is the Best Solution to Support our Employees? Paddy Mills HBR recently spoke with Stefan Bartscher, the head of innovation management technical planning at BMW, about the company’s pilot program deploying 20 adaptive robots at its plant in Spartanburg, South Carolina. Edited excerpts follow. How have your employees reacted to working alongside adaptive robots? In general the reactions have been very positive. The robots are taking a lot of difficult work away from our employees. And they help increase quality, flexibility, and process stability. To give you an example: Door sealants need to be applied with a continuous pressure of about seven kilograms. And employees need to do that in a very uncomfortable position. One can easily imagine that after several doors, precision drops. Therefore, we have given the job to a lightweight robot that can maintain precision and apply pressure continuously, without taking any breaks. The employee is happy, because he or she is relieved of an exhausting job and is able to take on something different. At the same time, we are assured of getting the same high quality in every car. Will your investment in adaptive robots grow, and if so, what will that mean for workers? We are quite sure that the number of lightweight robots in our plants will increase during the next years. But we need to keep in mind that solutions other than just implementing a new robot might be even better and more cost-efficient. The question is always: What is the best solution to support our employees? Often we conclude that in a certain area, robots would not make sense. Have you gained any unintended second-order benefits from the new generation of robots? For example, do they help you widen your hiring pool because they take on physical work that used to restrict whom you could hire for the factory floor? We would say that is one of our first-order benefits, especially with regard to demographics in Germany. Having lightweight robots work side by side with people on the line offers the possibility of “demography proofing” our aging society by providing ergonomic support. Shah is studying manufacturing facilities that have deployed adaptive robots, and her research suggests that four key benefits are emerging: 1. Robots have become safer, and this makes people happier with them. In one survey, workers said they felt safer and more comfortable with an adaptive robot than with a standard robot and that they were “more satisfied with it as a teammate.” 2. Robots speed up thankless tasks. Having an adaptive robot perform non-value-added work enabled employees in one study to complete tasks 25% faster than when they were assisted by a fixed robot. This is important, because such work often creates bottlenecks on assembly lines. Even a few seconds saved on non-value-added tasks can accrue significantly over time. Imagine that a factory makes 20,000 cars a month, each of which sells for $30,000. If it runs two six-hour shifts a day (allowing time for lunch and other breaks), that’s about $27,777 of revenue earned or lost every working minute of the month, or about $1.7 million an hour. If 12 minutes of every hour go toward non-value-added work, the 25% time savings from adaptive robots would cut non-value-added work time to nine minutes. The three minutes thus saved would translate to about $83,000 of potential revenue. That’s roughly $1 million during a two-shift day, or $30 million a month. 3. Robots greatly reduce idle time. In one of Shah’s studies, the switch from standard robots to adaptive ones allowed workers to complete tasks almost 6% faster, with 3% less idle human time and 17% less idle robot time. In another study, Shah found that people working with smart robots that could anticipate the need to gather building materials spent just 6.5 seconds idle, on average, during a 10- to 15-minute task—doing no work while they waited for the robots to recognize that it was time to retrieve and hand over the needed materials. People who had to ask the robots for materials spent 44 seconds idle. 4. Workers want to work with the new robots. Adaptive robots seem to be changing people’s perceptions of the machines’ value and potential, to the point where workers are willing to cede authority to them. In one experiment, Shah found that giving humans more control over allocating tasks—putting the people in charge—made them less likely to want to work with robots again and negatively influenced team fluency, or how well the group performed. In a sense, workers prefer to have adaptive robots allocating tasks. (See “When Your Boss Wears Metal Pants,” HBR, June 2015.) Huge gains will come when the new robots are paired with high performers. Other positive outcomes are emerging. Even smaller companies can invest in robots as the price of acquisition comes down, increasing opportunities for innovative uses. Because the new robots can move more freely, manufacturing spaces can become smaller and more flexible. And the makeup of the workforce can change: As robots take on more non-value-added work, older, less mobile employees can more effectively work on factory floors. What excites Shah most of all, though, is adaptive robots’ ability to learn from people. Specifically, she says, huge gains will come when the new machines are paired with high performers: “Oftentimes in factories, there are key players who are good at certain things, and if something goes wrong, they’re the ones who can reconfigure and fix the problem quickly.” When these high performers aren’t available, costly bottlenecks arise. It’s hard to reproduce a high performer in a machine, because so much of what makes the human effective comes from accumulated knowledge and an understanding of how things really work. But if robots can observe high performers and adapt their techniques, Shah says, they can provide even more benefits. “When we start porting knowledge from ‘high po’ talent to an AI engine, that’s huge,” she says. As adaptive robots increasingly gain the capacity to be trained by people and to be programmed on the fly to do whatever needs doing, their efficiency and flexibility will skyrocket. Already they have made manufacturing automation a less extreme commitment than it once was. In the future it may be hardly a commitment at all. About the Research: “Efficient Model Learning from Joint-Action Demonstrations for Human-Robot Collaborative Tasks,” by S. Nikolaidis, K. Gu, R. Ramakrishnan, and J.A. Shah

Smarter, Smaller, Safer Robots

Partner Center