Golem: an algorithm for robust experiment and process optimization

Area of Science:

• Optimization algorithms
• Autonomous experimental platforms
• Robust experimental design

Background:

• Science and engineering optimization tasks often assume precise, reproducible conditions.
• Existing strategies fail to account for variability, leading to suboptimal and unreproducible results.
• Optimized parameters may perform poorly under variable operating conditions.

Purpose of the Study:

• Introduce Golem, an algorithm for robust experiment and process optimization.
• Enable the identification of solutions resilient to input uncertainty.
• Ensure reproducible performance of experimental protocols and automated processes.

Main Methods:

• Golem is an algorithm agnostic to specific experiment planning strategies.
• It identifies optimal solutions robust against input uncertainty.
• Applicable for analyzing past experiment robustness or guiding real-time planning.

Main Results:

• Golem ensures reproducible performance of optimized protocols and processes.
• Demonstrated through extensive benchmark studies.
• Successfully optimized an analytical chemistry protocol under noisy conditions.

Conclusions:

• Golem provides a method for robust optimization in science and engineering.
• Enhances the reliability and reproducibility of experimental outcomes.
• Offers practical relevance for real-world applications with inherent variability.