It's widely believed that we're in the middle of a robotics revolution, but at this stage robots are still largely confined to cages doing tasks that don't require a lot of intelligence or interaction with us humans. We spoke with John Lizzi, Manager of the Distributed Intelligent Systems Laboratory at GE Global Research, about General Electric's approach to the future of robotics – specifically the future of what the company calls "service robotics," where robot apprentices will work closely with humans and take over many of the dull, dirty and dangerous jobs of today.
Gizmag: What is service robotics?
Lizzi: GE has historically made a lot of large industrial assets. We've been successful with that and continue to do so, but over the past few years we've been moving toward making a lot of money on the services that people need for those hardware assets. For example, there's our initiative around the industrial internet. It's focused on how we make these machines more connected, more intelligent, and robotics is an extension of this.
The concept of service robotics is how we can put autonomous or semiautonomous robots into [the work environment] to help our customers and deals with the "three D's" of robotics: dull, dirty, and dangerous. If you think about the environment where GE typically finds itself, the equipment is in places like factories, rail yards, refineries, and power plants, where dull, dirty and dangerous work is and where we see a lot of opportunity for robotics. This includes collaboration with humans and autonomous robotics that's an extension of human capabilities.
Is this a different sort of robotics from what we see now?
If you look at the practical applications of robotics, the vast majority of robots that you see today are in factories doing high precision, high speed work, such as sticking, placing, grinding, deburring, painting, that sort of thing. The first generation robots are extremely good for a lot of things, but they're not very aware of their environments; they're not adapted to work around humans. They're in cages, where they're separated from humans, and they're extremely expensive in terms of both the robots and the support equipment around them.
The new generation of robotics is riding a lot of trends, such as Moore's Law, so we can put more intelligence on the robot itself. The costs of computation and sensors are coming down. There's this whole movement around collaborative robotics with robots that are very easily taught, very cheap, and very able to work closely with humans, while employing new technology, such as SLAM (Simultaneous Localization And Mapping) for autonomous vehicles and similar applications. These are coming together and allow us to let the robots out of those cages and working in more dynamic, more unconstrained environments that I've mentioned.
Are we talking about a robotic apprentice that would be working alongside a more skilled human?
We see a range of different things. For example, there are applications around assembly. There are some tasks that humans are really good at and it will take some time to catch up with these. There's dexterity, manipulating small things, creative tasks in terms of assembly. We still want humans to do what they're better at, but there are others where robots could do a lot of work on things like going and grabbing parts, handing the person tools, transporting materials from one place to another. There's a lot of places where we can leverage the skills of both humans and robots.
You can also think of it as a contest of nonhuman skills. Imagine a larger robot that could give a human superhuman strength or imagine something large that GE makes that a human could use a robot to guide into place.
When we get to the more non-factory industry environments, such as power plants, you could have a robot being an apprentice to the human. The human and robot could start by working together and then over time the robot could start learning some things and that robot will start doing some of that more proactively, while the human focuses on other tasks.
There's also task coordination, such as repairing an asset and the robot could be collaborating with the human by providing extra physical or virtual capabilities.
The robots could also be an advance team. Instead of having engineers sitting out there all the time, there's no reason why we couldn't have the robot out there doing the work and being sort of the front for servicing the asset or taking a look at what happened. You could imagine a robot roving a power plant and having an understanding of what's normal and what's not. It could dynamically reach out to an engineer a thousand miles away and interact before we deploy a team.
What are the dirty jobs?
One would be inspection of electrical transformers and high tension lines where people get strapped up and go to extreme heights as part of a regular service. You could imagine an aerial-based system that takes over from the humans. Robots can start at the ground, go up high and inspect something. They can also operate horizontally and inspect miles of a pipeline or of a transmission system or what have you.
One of the things that we're particularly excited about are the core technologies across all these areas that we believe, once they're built, could be reused and reapplied; things like perception, navigation, and manipulation. We're building on these with our partners that will be critical to the future of service robotics. In addition, there's some work in our factories where we'll see collaborative robotics pop up dealing with repetitive and ergonomic stress-inducing applications.
What do you see as the big hurdle that needs to be overcome to create these service robots?
I think the big challenge is going to be in building more sophisticated robot perception, more accurate manipulation for the robot, and for how the robot can validate and verify autonomous systems. It's easy to do this where you can control the environment, but open this up and it becomes more difficult to validate and verify; it becomes more and more of a challenge.
You talked about service robots working closely with humans. What about safety? For example, there was that recent incident of a worker at Volkswagen being killed by a robot.
That was a traditional robot that was not aware of its environment and could do some major damage. Part of the trend is making robots that can sense when they come into contact with a human and comply with that. There's a lot of the hardware that can sense contact, but there's other ways using sensors and software to understand the objects around it that can address those issues.
When you address safety, you're going to give up speed and precision, which is a major tradeoff. A lot of the perceptive systems aren't as precise and don't have the payload capability that traditional robots have. There will be a trend where the capabilities of new robots will eventually meet those of traditional factory robots, but we're not quite there yet. There will need to be an evaluation of where you use a traditional robot versus this new generation of robots where you have robots that have a little less capability.
The recent changes are that these robots can somewhat work in an environment with humans, are smaller and not necessarily stationary and they can learn new (but still repetitive) task through 'observation'. That is a giant leap forward.
Going beyond controlled environments or repetitive tasks are still some ways off. Powering robots in uncontrolled environments requiring a lot of movement is still problematic due to the power requirements. I suspect not until we develop material with the ability of muscles (that can very efficiently relax and contract on small electrical impulses) would we be able to efficiently power robots to be able to work with humans at their speed, agility and strength (and possibly much stronger).
For now great strides are being made, but we're still in the very early days of the advances required to have the kind of robots that people dream of helping them around the house. For industrial purposes though, we are making great strides.