BY STEVE RANGER, EDITOR-IN-CHIEF | IMAGE: Johnny Andrews/UNC-Chapel Hill
Introducing robots and artificial intelligence (AI) into chemistry labs could help speed up the pace of scientific discovery by allowing experiments to be done faster and more consistently.
Researchers at the University of North Carolina at Chapel Hill, US, have been testing the use of robots in the lab and say that in future these autonomous assistants will help free up time for scientists by doing repetitive manual tasks for them instead.
When it comes to research into new molecules or materials, many projects follow a standard Design-Make-Test-Analyse loop. Scientists design a new material, then they make it, test it, analyse those test results and then, based on the findings, they can go back and update their original design to make a better material – and then start testing again.
This loop often goes around – and around.
‘That’s a lot of manual effort and what we’d like to do is bring robots into this, so that they can automate this making and testing. If we can do that, we can make and test things faster and do it more reproducibly and speed up that loop. The benefits are big,’ says Professor Ron Alterovitz of the Department of Computer Science at the University of North Carolina at Chapel Hill.
Robots can be useful in the lab because they will run experiments the same way each time and don’t need to take breaks.
At the moment the levels of automation in most chemistry labs are modest. Labs may have equipment that automates a single step, such as a gas chromatography machine or automatic pipetting machines. In contrast, full automation would see robots and AI completing an entire make-and-test routine from start to finish.
‘That’s very much science fiction today but we want to push the research in that direction. One of the big research directions we think can help with that is to have mobile robots in the environment that can help people do tasks, Alterovitz says.
The researchers have been working to add automation into their scientific workflow, such as through the use of robots that can go between lab stations with samples. One of the tasks they started with was needle injection, a common task for transferring liquid or gases. The labs are working on projects including solar panels and solar fuels. ‘In all these problems there’s just a huge chemical space to explore. Everything is variable so it’s a huge space to search and small changes can often lead to big improvements in the performance of materials. These robots and automation will allow us to search that space faster. The more of that space we can search the better solution we can find,’ he notes.
As well as working on increasing the skills of the robots in terms of picking and placing objects, the researchers are also keen to build out the variety of action commands to make the robots more general purpose. As these robots are not going to be working in an empty lab, there is also the question of how robots can safely and efficiently share the space - and the lab equipment - with humans.
The researchers have proposed five levels of laboratory automation in their paper Transforming Science Labs into Automated Factories of Discovery, published in Science Robotics.
Assistive Automation (A1) sees individual tasks, such as liquid handling, automated while humans handle most of the work.
Partial Automation (A2) sees robots perform multiple sequential steps, although humans are responsible for setup and supervision. Conditional Automation (A3) arrives when robots can manage entire experimental processes – with humans intervening if unexpected events arise. High Automation (A4) sees robots execute experiments independently and being able to react to unusual conditions. Full Automation (A5) is when robots and AI systems operate with complete autonomy, including self-maintenance and safety management.
AI plays a key role in advancing automation beyond physical tasks by analysing datasets generated by experiments to identify patterns and suggest new compounds or research directions. In AI-driven labs, the whole design-make-test-analyse loop could become fully autonomous, the researchers said.
Alterovitz says that, as measured by these levels, most labs today are likley to be operating at around A1 or A2 and while there may be some labs around A3, nobody has achieved A4 or A5. The highest level of automation is still potentially decades away; it would require the AI running the lab to be able to understand all the potential interactions of the ingredients being used in the lab to ensure safety.
Alterovitz says lab automation can reduce the need for humans to do the manual tasks in the lab, freeing up scientists to work on the higher-level chemistry problems that need to be solved – and setting up the best experiments.
‘There are these higher-level aspects of the research that humans will be able to spend more of their time on and the effect is you can make more science progress faster,’ he adds. As well as making sure the robots can work well with humans, training scientists to work with advanced automation systems will also be important.
Scientists will need to understand the capabilities of robots, data science and AI to accelerate their research. ‘Educating the next generation of scientists to collaborate with engineers and computer scientists will be essential for realising the full potential of automated laboratories,’ the researchers say.
Visit Blog
