Study: Laboratories face profound change through robot automation

A research team at the University of North Carolina Chapel Hill (UNC Chapel Hill) believes that scientific laboratories are facing a profound change triggered by robot automation and artificial intelligence (AI). They could be used to speed up scientific experiments and make them more precise. As a result, scientists expect more and faster breakthroughs in the fields of health, energy and electronics.

"Today, the development of new molecules, materials and chemical systems requires intensive human input," says Dr. Ron Alterovitz, professor in the Department of Computer Science at UNC Chapel Hill and lead author of the study "Transforming science labs into automated factories of discovery", which was published in Science Robotics. "Scientists need to design experiments, synthesize materials, analyze the results and repeat the process until the desired properties are achieved."

Leaving standard laboratory tasks to robots

This "trial and error" approach takes a lot of time and is very labor-intensive. The automation of these processes by robots and the use of AI is therefore a logical conclusion. Robots can carry out experiments continuously without tiring. This would speed up research considerably, the scientists write.

Robots can work much more precisely than humans and also carry out the steps of individual experiments more precisely and consistently. The safety risks are also lower, for example when dangerous substances have to be handled. The robots would relieve researchers of routine tasks, giving them more time to concentrate on more important, higher-level research questions. This would make it possible to achieve faster breakthroughs in different disciplines.

However, it is not quite that simple, the research team admits, as this requires the development of robots that can work together with the scientists in a shared laboratory environment. This requires "creative solutions", the researchers write.

AIs must also be developed to analyze data sets generated by experiments, recognize patterns in them and provide new ideas for further research. For example, they could also autonomously determine which experiments are carried out, make adjustments to the experiments in real time and thereby improve the research process. Together with robots, almost the entire laboratory operation could be automated.

"We expect that with continued development, robotics and automation will improve the speed, precision and reproducibility of experiments across different instruments and disciplines and generate data that can be analyzed by artificial intelligence systems to guide further experiments."

Five stages of laboratory automation

To achieve this, the scientists have developed five stages for successful laboratory automation in their study:

• Assistive Automation (A1): At this stage, only individual tasks are automated, such as handling liquids. The majority of the work is still carried out by humans.
• Partial automation (A2): In the second stage, robots already carry out several consecutive work steps. The researcher then only has to set up and monitor the automated processes.
• Conditional automation (A3): Robots take over the entire experimental procedure. Humans only intervene in the event of unexpected events.
• High automation (A4): Robots take over the execution of experiments independently. They also set up the necessary equipment and react to unforeseen events without human intervention.
• Complete automation (A5): At the highest level, the robots and AI systems work completely autonomously. They can then also maintain themselves and manage their own safety.

These levels can serve as a guide for gradually automating laboratories in the future, the scientists write.

However, the implementation of automation is technically and logistically challenging. This is because laboratories have different requirements depending on their type, from individual process laboratories to large-scale laboratories. The automation systems must therefore be designed flexibly, for example using robots that are mobile and can carry out tasks at different stations.

In addition, the scientists must be specially trained to handle the systems. They need knowledge of robotics, data science and AI that goes beyond their specialist discipline. Only then can the potential of automation be fully exploited, according to the researchers.

(olb)