Product configuration is a successful AI application which operates by selecting and assigning components from a catalog in accordance with the customer and configuration requirements.
Configuration problems occur frequently during configuration of telephone switching systems, electronic railway interlocking systems, automation systems, etc. Often such problems are subject to complex optimization preferring solutions which include a minimal number of components, thus, minimizing the overall production costs.
A reconfiguration problem arises when a product or service is not designed from scratch, but has parts of the existing configuration adapted. This is an important activity in the after-sale lifecycle for companies selling configurable products or services. Typically such products have a long lifetime and their configuration requirements are changing in parallel with the customers' business.
The development of knowledge-based (re)configuration systems requires the application of a knowledge representation language which on the one hand is expressive enough to capture a (re)configuration model, and on the other, there should exist reasoning methods for it that allow a solution be computed efficiently. This PhD project focuses on the development of a knowledge representation language which allows encoding of different (re)configuration problems occurring in practice of Siemens and in general. The reasoning is done by translating the knowledge base to modern formalisms such as Answer Set Programming and Constraint Programming.
In practice finding a solution for (re)configuration problems using general frameworks might result in an unacceptable performance, because of the large number of components and the presence of symmetrical solutions. Therefore, a decomposition method was developed that allows partitioning of these problems into loosely coupled sub-problems for which solutions can be computed more easily than a solution to the whole problem. While preserving all solutions, the method significantly improves performance and quality of solutions for real world configuration problems.