StarFLIP++

The goal of real-world decision making is to optimize some combinatorial problem. In order to optimize, one needs criteria, which can be expressed for instance as constraints. Real world constraints are usually vague and have different relative importances. Also, data collected from unreliable sources is often slightly incorrect. Sometimes criteria even contradict each other. Therefore compromises are necessary. Typical application areas of combinatorial constraint optimizers include scheduling, design, configuration, planning, and classification. Test application domains for prototypes are steel-production scheduling and classroom roostering. To investigate real-world combinatorial optimization problems, we have designed and partly implemented the StarFLIP++ family of C++ libraries, especially focusing on reusability and applicability for a wide range of combinatorial problems.

StarFLIP++ will ultimately be composed of the following layered sub-libraries:

• the fuzzy logic inference processor library, FLIP++
• the fuzzy constraint library, ConFLIP++
• the dynamic constraint generation library, DynaFLIP++
• the domain knowledge representation library DomFLIP++
• several heuristic optimizing libraries (currently tabu search and genetic algorithm), OptiFLIP++
• the userinterface, InterFLIP++, with platform support for X-Windows (XView/Open Look, Motif) and MS Windows 3.1/95/NT
• the HTML documentation (end-user/knowledge-engineer/programmer), DocuFLIP++
• the knowledge-change consistency checker library, CheckFLIP++
• the version-control and test environment for the complete library set, TestFLIP++
• the simulation toolkit, SimFLIP++
• the reactive optimizer, ReaFLIP++
• the neural network extension that allows automatic tuning of fuzzy membership functions, NeuroFLIP++

 2007–2009 Alexey Beshenov.
Contact: al@cadadr.org.