New industrial robotics applications are enabled by improvements in artificial intelligence (AI), the potential of 5G wireless, the Internet of Things (IoT), and cloud computing. However, the next wave of industrial robotics also requires new skills and training for the humans who will design for and work with them.
The automotive sector has long pioneered industrial automation, but now the practice is expanding and flourishing in other areas. Improvements in enabling technologies such as computing, sensors, and machine vision—and some advances in materials—have helped collaborative and mobile robots take hold in industries such as logistics. That is where the growth opportunities exist. This subgroup of collaborative robots is changing the way people work with machines because they are designed to work alongside people. Furthermore, they tend to be much cheaper than traditional automation.
Where robots lead the way
The automotive industry is venturing into designing automation and human-assist technologies that depend on sensors and AI. Companies are spinning off centers, such as the Toyota Research Institute, to develop technologies in other areas such as human factors and self-driving cars (see “Case Study: Automating for Process”).
“Industrywide, the biggest excitement is about how self-driving car technology can be used in cars and trucks, airplanes and ships,” says Vijay Kumar, Ph.D., Nemirovsky Family Dean of Penn Engineering at the University of Pennsylvania. In the university’s GRASP Lab, Kumar and his students create autonomous ground and aerial robots and design bio-inspired algorithms for collective behaviors and robot swarms.
Kumar sees growth in data analytics systems in the logistics field, such as advances in sensing, communication, and storage technologies that enable acquiring and storing data “at unbelievable levels of resolution,” he says. He adds that advances in data production, AI system integration and availability of sensor data for input into computing systems “enable sensing, computing, storage, and communication with the ability to mine the data, understand it, and to learn from it.”
While actual robots may not have changed that much in the past 10 years, Sean Murray, co-founder and director of robotics engineering, Realtime Robotics, a startup with the goal of expanding the potential of robotic automation, says there have been improvements in 3D sensors, machine learning algorithms to interpret data, as well as mobile robotic applications.
Murray adds that for anything that is difficult to deploy, there is a technology learning curve. Scale and scope matter.
“It only makes sense to invest in acquiring that expertise if you’re going to amortize that cost over a lot of factories or a lot of deployments,” Murray says. “While larger companies are early adopters, as automation technologies become easier to use, smaller and lower-volume industries can take advantage of innovations as well.”
“Industrial robotics prospers where you have the right intersection of volume and value,” says Erik Nieves, CEO and co-founder of Plus One Robotics, which develops 3D vision software and collaborative automation for the logistics industry. “And that’s why automotive is still the dominant market for industrial robots. Historically, when you have repeatability, robots win. It can’t just be a million widgets; it’s got to be a million widgets of substantive value.”
According to Nieves, the burgeoning online commerce and logistics industry is a prime area for industrial automation in the United States. “We are moving more of our commerce online,” he says. “Which means that supply chain has a growing labor problem because increasing order volumes and shorter delivery cycles require a huge amount of labor.”
The needs of the current supply chain industry harken back to the original impetus for the development of robots. “Remember the classic three D's of robotics? Work that’s dirty, dangerous, or degrading to the human spirit? In supply chain, there’s a fair amount of that,” Nieves says. “It’s not very dirty, but it is certainly dangerous from the standpoint of ergonomics. Nobody wants to move stuff from left to right, every two seconds for the next six to eight hours. That’s why the churn is so bad.”
Although robots excel at high-value tasks of low variability, a manufacturing environment usually still ends up with more people than robots. “That’s even true in automotive because you have repeatability and a constrained set of processes on the 'body in white' side of the wall,” Nieves says. “But when it crosses over to the final assembly, now you have a lot more variability in trim levels and options.”
However, once a job requires more cognitive complexity, robots can lose their edge. “Robots are great because they’ll pick up and sling 250 pounds all day long,” Nieves says. “But they’re dumber than a box of rocks. Anytime a robot needs to make a qualitative decision about its environment or circumstance, it breaks down. People are just better.”
The latest improvements
“Systems that have moved around warehouses have been around for a long time,” says Thomas Ryden, executive director of Mass Robotics, a nonprofit supporting the Massachusetts robotics community. Now, startups are developing flexible, intelligent systems that can quickly plan on the fly and gather data as they do, as well as fleets of robots that can navigate and communicate with each other to optimize performance.
“Instead of the robot going from point A to point B to point C, a system of collaborative robots can communicate and select the closest robot who has the capacity, who can get back to the pick stations fastest,” Ryden says. “And then, use cloud-based IoT and AI to optimize for mobile manipulation. So, think of adding traditional factory automation arms—but now on a mobile platform. You would want to avoid collisions with other robots and people in terms of the trajectory of the arm or arms.”
These collaborative robots can work with humans in the same workspace because the robots can detect human movement and then plan a dynamic path and execute tasks in close to real time by using field programmable gate arrays, a hardware-based solution. “And so, this whole idea is of computing at the edge—putting the processing right next to the arm with a combination of sensors and sensor fusion to be able to accurately position and reposition the arm so the path that it takes avoids all obstacles in the area,” Ryden says.
He also points out the new trend toward leveraging AI and other sensing capabilities to ease the programming load. In traditional factory automation, an engineer has to program a robot to carry out a specific task, but now this can be accomplished more flexibly through software.
Automated guided vehicle (AGV) technology has been around for about 25–30 years, according to Melonee Wise, CEO of Fetch Robotics, which develops robotics for logistics. The reason autonomous mobile robots have become so popular is that they offer flexibility and the ability to deploy on demand in a dynamic environment. Wise believes that technologies such as the cloud and potentially 5G—as well as new, low-cost sensors—will influence industrial robotics development. Expect to see autonomous mobile robots with arms on them, as the industry is already starting to prepare for this by developing standards.
Meanwhile, Nieves speculates that the future of the robotics space is not autonomous robots, per se, but robots which exert some degree of local autonomy while aided by a human. That human assistance is necessary for the edge cases—the unexpected events which fall outside of a robot’s programed capabilities. In these situations, a human must step into the process to make the decision.
“With this work model, edge computing happens at the work cell, but there’s still a human in the loop,” Nieves says. “And that human can be anywhere there is internet access.” Connectivity to the robot in the field has now become important, and while the tools and infrastructures exist that can connect these robots using cloud computing platforms, cybersecurity remains an issue.
Nieves believes that future industrial robots will work autonomously 80% of the time—a minimum number he believes is required to justify the return on investment. While one may not hear from a specific robot for hours, when it requires decision assistance, it will message a human to provide direction on the next course of action.
The logistics field needs further development of technologies in perception and grasping, as well as bilateral and mobile manipulation, according to Nieves. While manipulating a robotic arm is a straightforward process, other technologies under development include scene segmentation, 3D manipulation, decision-making, and the ability to iterate quickly on designs with additive manufacturing.
There is also a need for better simulation tools to train robots. “It used to be in the old, line-building days for automotive, the best you could do is actually get a physical robot, put your tool on it, program it, and see what cycle time you’ve achieved,” Nieves says. “We’ve gotten smarter about how to replicate these environments digitally. While many robot manufacturers have proprietary simulation tools, there are also third-party robot-agnostic simulation tools available. These third-party tools mean you can program any robot, but these simulation packages are often more expensive."
Working relationships with robots
According to Murray, in deploying industrial automation, the first challenge is to allay workers’ concerns that they must deal with maintaining a piece of equipment that breaks frequently. “After six months, they uninstall the factory automation, and people are doing the job again,” he says. “And then on the management side, same thing. The challenge also is convincing them that you’re going to get them a return on their investment.”
Fetch Robotics emphasizes the user experience to help ease concerns, too, Wise says. “We started to mimic a lot of the interfaces that you see on the internet and web browsers, so that people feel very comfortable with using and configuring the robots,” she says.
Her company also developed guidance, called “robot etiquette,” to help users interact with robots safely. “You should treat the robot like another co-worker,” Wise says. “You wouldn’t step out in front of a co-worker, jump in front of a co-worker, push a co-worker, or drag co-worker around.”
The other key is training the workforce at all levels, and getting employees to a place where they can operate these robots comfortably.
“Typically, the fear that workers have is not that they’re going to lose their job; it’s that they they’re not capable or competent to work with the robots.” Wise says. “You have to remember that many of the people that are working with our robots don’t have a lot of formal secondary education, and so they are constantly worried that they can’t keep up. That’s why we’ve invested so much effort into making the robots easy to use. A lot of what we built into the robot, like process engineering a warehouse, is really like visual programming. So, it’s a pretty easy-to-use way to configure what the robot has to do and how it’s supposed to do its job. But that depends on the person that’s using it.”
For engineers and developers, robotics is as much or more software development than pure-play engineering anymore, according to Nieves. He advises younger engineers to add software and programming language classes such as Python, C++, or Robot Operating System (an open source robotics framework) to their curriculum because so much of their future depends on being able to code.
However, Nieves also wants potential hires to have a good background in human factors. “It’s not enough to be a whiz-bang software developer and great at controls,” he says. “And you know, maybe you can even engineer the mechanicals. But if you don’t understand what the person who does this task experiences each and every day, then you’re not going to have sufficient empathy to engineer the best solution for the worker. And that’s the one thing I ask interview candidates here at Plus One: Can you really relate to the people who will be helped by our robots?”
Written by Wendy Wolfson