“The ultimate dream is to accelerate the discovery of new materials for the energy transition by combining artificial intelligence, robotics and data-driven experimentation in a self-driving lab.” This is how Maas van Apeldoorn and Delaram Khalili Samani describe the ambition of DIFFER, which has made a flying start with the recently completed final projects of the two technical doctorates.
The two-year post-master Engineering Doctorate in Software Engineering is designed to prepare a trainee for an industrial career as a technological designer, and later on as a software or system architect. The program is comprised of 14 months of advanced training and education, including 3 industry driven group projects, followed by a major individual 10-month industrial graduation project.
For their final graduation projects, Maas van Apeldoorn and Delaram Khalili Samani decided to embark upon DIFFER’s journey toward realising a self-driving lab to accelerate renewable energy research. A self-driving lab is an automated laboratory that autonomously runs experiments – including the design, synthesis, and testing of materials – in a continuous loop, where the results of one experiment inform the design of the next.
Become a better engineer
“When I first read the project description, I was immediately intrigued by the challenge of building such a lab”, says Samani. With her background in computer engineering, she first gained some work experience as a freelance back-end developer before she decided she wanted to advance her knowledge and expertise. “Originally, I was looking for a PhD position, when I saw an advertisement for this EngD program. Since I did not aspire an academic career but wanted to end up in industry as a software architect, I was intrigued by this dedicated form of post-master education. I like the fact that as an EngD trainee you are trained in ‘soft’ skills like project management, and that you get the opportunity to directly work with industrial clients.”
Her colleague Van Apeldoorn ended up at DIFFER as a result of his deeply felt wish to do something meaningful. “I have always known that I do not want to work for some high tech company just to earn a lot of money. I chose to enroll in an EngD program to gain some practical experience and combine that with personal development aspects, to find out where I fit in best and can be the most effective. After my first meeting with Murat Sorkun, project leader of the DIFFER self-driving lab digital infrastructure, I knew I had found the right assignment for me. Besides the personal click I experienced with Murat, I felt I could really make a tangible difference here.”
Integrating stand-alone laboratory devices
Currently, DIFFER’s materials discovery lab consists of a wide variety of manually operated devices, ranging from material deposition systems to electron microscopes, Samani explains. “Measurements are often stored on a USB stick and then transported to a researcher’s individual computer for further analysis. I have been working on connecting these devices to a software platform that automatically collects the output and stores it in a uniform way. The problem is that we are dealing with different types of machines, built by different vendors. Each device has its own data format, networking technologies, and characteristics.”
Given that most of the devices are in active use by researchers, most of the time, the EngD candidate could not work directly on the actual machines. “That is why I built emulators for all machines, replicating their hardware and software. Doing this, we can develop the system without occupying the machine or making potentially expensive mistakes”, Samani explains. At the end of her project, four of these emulated devices have been connected to the centralised software platform she developed, and one machine is physically integrated in the software.
Design your own lab
Where Samani focused on the physical integration of the different lab set-ups, Apeldoorn worked on the opposite side of the spectrum, he explains. “At the moment, there is little to no tooling to design these types of self-driving labs before you start investing in the required machinery. I developed a simulation framework to test different laboratory configurations in a virtual environment. My framework models self-driving labs based on four entities: machines, employees, robots, and experimental workflows. We designed the framework to be domain-agnostic, so it can be used across many scientific domains and not just electrochemistry. Through this tool, users can get some insights in the lab’s main KPIs, like the amount of experiments per unit of time, projected costs and device utilization.” At the moment, the system model is text based, but the aim is to turn it into an easily accessible drag and drop system that non-computer experts can use to design their own lab in a much more intuitive way.
“The challenge was to first understand the domain DIFFER is working in; I do not know anything about electrochemistry”, Apeldoorn says with a smile. “The next challenge was to come up with a framework that is both reliable and still flexible enough for any device to be added, for any domain.”
Large steps
Looking back, both recently graduated EngDs are positively surprised with what they managed to achieve in a mere ten months. Apeldoorn: “When we started back in January, the proposal to build a self-driving lab at DIFFER was just three months old and the project team was just the two of us, one experimentalist and our project manager Murat Sorkun. As a team, we managed to produce more than anyone had expected upfront.” “Indeed, the key to success has been that we had a great team, with a lot of mutual respect and trust”, adds Samani.
Though both engineers have finished their projects and obtained their EngD title, they like to continue working at DIFFER. Samani: “I am staying on as a software engineer at DIFFER, and Maas is also interested in continuing.” Apeldoorn: “During our EngD projects, we experienced a lot of freedom to build what we think is best for this application. That is great and scary at the same time, since you bear a lot of responsibility. Murat has been an excellent supervisor in that regard, taking us seriously as engineers and respecting our expertise.”
All in all, the two EngD projects have established a shared trust that realising a fully automated lab is actually feasible. Apeldoorn: “We are now developing a project plan for the coming five years.” Samani: “It is nice to see how other departments in the institute are now also starting to recognize the potential advantages of what we are doing here.”
The two EngD projects are carried out in the early stage of an ambitious self-driving lab program at DIFFER. The goal of this large-scale program is to enhance all of the institute’s chemical energy laboratories at DIFFER through digitalization, automation, and AI, creating a high-tech environment that accelerates material discovery. To this end, two new research groups will be established early 2026. The self-driving lab ambition fits well within the institute's mission to perform leading fundamental research on processes, materials, and systems for the global energy transition, in close collaboration with academia and industry.
Author: Sonja Knols
Go to the News page.