Strategic Patterns
The science use case strategic patterns define high-level solution methods using experiment control architecture features at a very coarse granularity. Their descriptions are deliberately abstract to enable architects to reason about the overall organization of the used techniques and their implications on the full system design. The Strategic Pattern Catalog defines the following strategic patterns:
- Experiment Control
Certain predetermined actions need to be performed while running an experiment. This pattern would be used in all automated experiments that do not have feedback for steering the ongoing or designing the next experiment. Since autonomous operation requires to first figure out automation, this pattern provides a basic solution that covers most experiments performed at this point.
- Experiment Steering
Certain predetermined actions need to be performed while running an experiment to positively influence experiment progress. This pattern involves feedback for the ongoing experiment as an extension to Experiment Control. It offers autonomous operation and is used in experiments that require live feedback to adjust parameters.
- Design of Experiments
Certain predetermined actions need to be performed to run a set of similar experiments with different experiment plan parameters, depending on (prior) experiment results. This pattern makes use of either Experiment Control or Experiment Steering and additionally offers feedback between experiments, typically to define the parameters of the next experiment or next series of experiments. It is typically used in conjunction with probabilistic (e.g., Bayesian) or domain science based (e.g., physics informed) analysis of experiment results. This pattern is predominantly used in large-scale parameter studies, such as to find the optimal conditions of a chemical catalysis.
- Multi-Experiment Workflow
Certain predetermined actions need to be performed to run a set of experiments in serial (one after another) and/or in parallel (simultaneously). This pattern utilizes the other 3 patterns to orchestrate multiple experiments that may depend on each other. An example use case is the creation of a certain material using physical and/or chemical processes (e.g., catalysis) and the analysis of the properties of the created material in multiple experiments (e.g., spectroscopy and stress testing).
The features of these science use case strategic patterns and their relationships are compared in Table 1.
Feature |
||||
|---|---|---|---|---|
# of experiments |
1 |
1 |
Multiple |
Multiple |
Control type |
Open loop |
Closed loop |
Closed loop |
Open loop |
Operation type |
Automated |
Autonomous |
Autonomous |
Automated |
Extends |
||||
Uses |
||||
May also use or use instead |
Strategic Pattern Catalog