CASTRO is a Python tool designed to facilitate experimental design space exploration under various constraints. CASTRO is particularly useful when dealing with mixture and synthesis constraints or when additional experiments must be planned within a limited budget, or previously collected data should be exploited.

Key features include:

• Novel constrained sampling method that leverages Latin Hypercube Sampling (LHS) and Latin Hypercube Sampling with Multidimensional Uniformity (LHSMDU) in its core.
• A divide-and-conquer approach for complex design spaces with more than four variables, dividing the problem into manageable subproblems and reassembling solutions.
• Distance-based selection method for choosing next experiments to maximise information gain by selecting points furthest from existing data
• Quantitative design quality metrics: central and wrap-around discrepancy, and variance
• Supports integration of previously collected data and budget-limited experiment selection
• Integrated visualisation tools for insightful result analysis.

CASTRO includes two case studies for material composition design.