This section is a compilation of rough guidelines to model problems for the Multi-Agent Optimization Engine. They have been identified through experimentations and explorations with the engine as well as work with customers. Keep in mind that they may be true most of the time but false for a given problem. This section is meant to evolve as we learn more by working on more customer's problems and as the Optimization Engine improves. As the dust settles, it should be expanded into a proper Optimization Engine Modeling documentation and a basis for Configuration Engineers training.
At the beginning of the modeling process, the main questions to ask yourself are, in no particular order:
- What are the dimensions of my problem?
- What are the spaces and scopes of my problem?
- What are my levers, and where do my criteria apply?
- At what granularity levels are we working? (e.g. products? product families?)
- Which types of criteria should I use?
- How should I prioritize them?
The answers to these questions have a great impact on the structure of the multi-agent system and its ability to efficiently solve the problem. The next sections attempt to give insights as to how to approach these questions. Reading the Agents section in the Multi-Agent Optimization Engine page may be helpful to fully understand them.