A research team from The Hong Kong University of Science and Technology (HKUST) has developed a new modeling and optimization framework that helps researchers design modularized autonomous experimentation (MAE) platforms for new materials by determining the optimal combination of equipment before construction begins, thereby reducing both time and cost before the platform is physically built.
The study introduces a hybrid-automata-inspired approach that systematically translates experimental procedures into optimal equipment configurations. It was published in IEEE Transactions on Automation Science and Engineering under the title "Toward a High-Throughput Automated Materials Experimentation Platform: Hybrid-Automata-Inspired Modeling and Equipment Configuration Optimization". Prof. YANG Jinglei, Professor of the Department of Mechanical and Aerospace Engineering, and Prof. DUAN Molong, Assistant Professor of the Department of Mechanical and Aerospace Engineering, serve as the co-corresponding authors. Current PhD student MA Guoxiong and MPhil graduate CUI Haozhe are the co-first authors.
High-throughput modularized autonomous experimentation is transforming materials research. By automating sample preparation, synthesis, characterization, and screening, it accelerates discovery while improving repeatability and producing born-structured datasets, that is, parameterized data recorded in a fully structured format during experimental execution. The key contribution of this study is a design-stage method for equipment configuration. It deliberately avoids grand claims about transforming materials research; instead, it solves a narrower, concrete problem: giving platform planners a systematic, repeatable basis for equipment-count decisions before hardware is purchased. Its value is that a methodology exists before the platform is assembled, so early design choices are not left to intuition alone.
To tackle this, the team built a modeling framework inspired by hybrid automata, a formalism for describing systems that move through discrete states over time. Experimental procedures are broken down into procedure states, transitions, timing logic, and equipment involvement. A structured equipment dictionary captures each device's function, operating parameters, cost, footprint, and dependencies on other instruments.
Using this unified model, the equipment configuration problem is formulated as a constrained integer nonlinear optimization problem. The formulation incorporates multiple design constraints, including budget limits, platform footprint, throughput targets, and inter-equipment dependencies. An integer-coded differential evolution algorithm then efficiently explores the discrete decision space to find feasible, cost-effective configurations.
The method was demonstrated on a thermal insulation coating case study conducted on a 0.54 m² desktop lab under an approximate US$11,000 budget. In a three-sample batch comparing automated against manual operation, active operator time fell by 79.1% (from 1,589.7 seconds to 331.8 seconds). Furthermore, the coating's light transmittance showed a smaller standard deviation, demonstrating improved repeatability and process consistency.
Prof. Yang noted, "This work provides a systematic approach for transforming traditional manual experimental procedures into a structured equipment-configuration decision-making mechanism. Instead of relying on intuition or experience alone, researchers and engineers can now analyze workflow timing, equipment utilization, and practical constraints at the early design stage - before committing resources."
Prof. Duan said, "Our goal is to shift equipment planning from experience-driven guesswork to quantitative, procedure-driven design. The framework helps identify equipment bottlenecks before they occur, improves resource utilization, and guides the construction of compact, high-throughput automated platforms."
The team frames this paper as the entry point of a series. Future efforts will focus on layout-aware configuration, dynamic scheduling, closed-loop AI-driven experimental decision-making, alongside extension to additional case studies and an end-to-end wall-clock throughput target. From a longer-term perspective, this work not only serves as an important foundation for advancing modularized autonomous experimentation but also represents a significant step toward realizing the team's vision of a Materials Operating System (MaterOS).
