As capabilities in artificial intelligence (AI) accelerate rapidly, researchers in the fields of machine learning and robotics have a lot to be excited about. For those at the National Robotarium - a new research centre for robotics and AI based at Heriot-Watt University’s Edinburgh campus - this excitement is fuelling projects aiming to support development of cutting-edge robotics that could provide disruptive solutions to a host of industries.
Alongside developing socially assistive robots and exploring how to improve human-robot interaction, another key area of research is Precision Laser Applications (PLA). Dr Richard Carter, associate professor in Heriot-Watt’s School of Engineering & Physical Sciences, is the academic lead for the PLA lab at the Robotarium. The laser lab moved into the £22.4m purpose-built facility following its official opening in September last year.
“We’re really interested in seeing how we can use lasers to make better robots,” Dr Carter told The Engineer. “There’s a really big interest in healthcare technologies for the kind of micro-robotics that can go inside people and do jobs … We have some projects we’re running to have a look at trying to manufacture things on that sort of scale, and out of the right materials, because you can’t just put any old material inside the human body – it’s got to be biocompatible, it’s got to be safe. These are all processes that take a bit of time.”
Dr Richard Carter, associate professor in Heriot-Watt’s School of Engineering & Physical Sciences
The lab is also interested in seeing how lasers can be useful for robots to carry out their tasks, such as in environmental sensing. A Heriot-Watt Professor, Derryck Reid, has created a new LiDAR system with higher resolution than standard systems, said to be capable of capturing data at hundreds-of-nanometre precisions, compared with the few centimetres normally possible with LiDAR. Reid is now working with project partner Renishaw (see interview, page 20) on advancing this technology, which could be game-changing in allowing for more accurate measurements in wide-ranging industrial applications.
“That gives us the opportunity to then strap that to a robot to do site survey work with much higher accuracy than what has been the case previously, or for these types of robots you see in factories doing precision work – you get this independent check of whether or not the robot’s doing it right, because you can measure exactly where it is in real-time and keep track of these things,” Carter said.
Using lasers as a tool for robots also feeds into medical work, where the lab is developing new processes, Carter explained. One of these projects, dubbed ‘PrECisE’ (Precision laser scalpel for cancer diagnostics and eradication), explores using short pulse lasers for surgical procedures.
“The advantage of using the short pulse lasers is you can very carefully control just how much tissue you’re taking away, to a much better rate than what would have been doing with traditional electric and scalpel style tools,” said Carter.
“We have to deliver the laser beam into the patient somehow and that’s where the robotics come in, it’s the endoscopic stuff. And the really exciting thing of course is to use the laser to manufacture the robotic tool, that then delivers the same laser into the patient for the surgical application.”
So, while in one case lasers can help robots to do their job, the lab is also interested in using the robots to manufacture the lasers themselves, he added, highlighting the scope of benefits from compounding advancements in robotic and laser technology.
“Manufacturing a laser is actually quite complicated, a lot of design and engineering and cost goes into trying to get something that should work on paper,” he said.
“And then of course you need to manufacture it, and not just one, you need to set up a production line to manufacture these things. One of the main problems is they’ve got very tight tolerances; you put these lenses, mirrors and all sorts of optical components down, and they need to be in the right place to fractions of a millimetre or it won’t work.”
Explaining this further, he said that time-consuming alignment processes have to be very precise – usually involving making small adjustments, imaging the laser beam and adjusting until it looks just right, a process that can be ‘exceedingly boring’ for the person carrying it out, he added. Additionally, it requires a fair amount of expertise, skill and training.
Whilst getting a robot to replicate the motion of a human has been well established, Carter pointed out, getting a robot to replicate the experience and thought process of how to go about building a laser is a ‘much more interesting challenge’.
An area his team is exploring in relation to this is human behaviour analysis, looking to see the human thought processes involved and whether a machine learning process can mimic it or get an equivalent result.
The lab sees a high level of collaboration with laser making companies, such as Luxinar, and developers of laser systems such as Oxford Lasers. Then there’s the industry end users, who are interested in using laser-based processes “but not really in knowing about the nuts and bolts,” Carter explained.
Several robotics start-ups have also set up their presence in the National Robotarium. Its tenants include Bioliberty, an assistive robotics company currently developing a soft robotic glove that could restore upper limb mobility in patients following a stroke; TouchLab, which is developing an innovative e-skin using its own sensing technology; and Crover, an agricultural robotics company that has developed a robotic solution for grain storage monitoring.
“We’re working with some of the machine learning [researchers] at the Robotarium to understand how we can apply that to help the laser processes irrespective of robotics – everyone wants to know what they can do with AI,” Carter said.
“I think the biggest challenge for robotics at the moment is the trust and regulatory bit…there needs to be a framework for how these things are going to be implemented, particularly as you start pushing robots out of industrial environments, out of the laser manufacturing labs, and into peoples’ homes and schools and hospitals, in areas where they’re interacting with the general public.”
The public concern that robotics and AI could go ‘beyond control’ is of course one that needs to be addressed, he stressed, adding that half of the work around human-robot interaction is convincing people that it is safe.
“We’re seeing massive strides in AI and machine learning over the last few years…the capability to do the decision making process is really coming along,” he said. “You can look at the history of chess playing robots – it’s well past the point where any human’s got any chance of not only being able to beat the robot but actually understanding what the robot’s doing. This is something that’s definitely going to be possible for a whole variety of areas.
“People aren’t going to adopt robots for a particular job if they aren’t aware that it’s possible. So actually telling people that it’s possible is nine-tenths of the battle.”