Rule-based Product Confi guration

Ching Hoo Tang & Christoph Weidenbach

Rule-based Product Configuration

In today’s industry and economy, the principle of product lines plays a great role. In this context, different products, based on a preferably high number of identical components, are grouped together and jointly designed, developed and produced. A typical example is the group-internal production of different automobile models on a common base.

A currently popular approach for describing product lines is the use of rules: They specify which components in a specific context, or following an explicit pre-selection of properties, can be configured into an overall product. These con- texts can be e.g. of a temporal or technical nature. Alternatively or additionally, a product can be described based on the hierarchical structure of its assembly. An example of a component selection for a vehicle would be the selection of a color, an engine or an equipment detail such as an air conditioner. A temporal context would include a particular year of manufacturing or model year; a technical context could be the upper limit of CO2 emissions or the total weight. The rules themselves describe dependencies between the components and contexts: A powerful engine requires specific tire combinations based on the potential achievement of high speeds, whereas trailer hitching devices exclude weak engines. The interactions represent so- called if-then rules, which become active through decisions by the user specifying properties of an eventual product.

Within the framework of our re- search we are interested in a formal description and verification of such a system of rules. We also want to be able to calculate mathematically precise analyses of the whole product structure. For many of the systems currently in use, the challenge is not the development of the concrete rules, but rather ensuring important global properties such as consistency. The above-described rule approach is particularly characterized by its flexibility. Rules can easily be adapted to the type of products being configured and the associated components and the respective area of use. The desired level of detail can be chosen freely: Thus, a product can be considered purely from a sales perspective, or down to the last, single screw for engineering purposes. The approach can also be applied to the description of products composed of abstract modules. In addition to the automotive industry, e.g. applications for software products that are composed of individual modules are also possible. Production lines, which consist of a sequence of individual steps, or the com- position of business services from individual activities, can also be naturally described with the aid of rule systems.

Although promising results have been accomplished with rule systems on the application side, an appropriate formalization, i.e. an explicit and consistent representation of a rule-based system by a language of logic, is still the subject of research. However, formalization is a prerequisite for the assurance of global properties. Important properties include consistency, i.e. consistency of the system of rules, or the uniqueness of the calculated results. Further examples of these properties are the total number of possible products from one product line according to the pre-selection of components or contexts, optimality properties of products such as minimum production costs, or recognizing components that no longer contribute to an active product. For systems of rules that also describe a particular process, e.g., a production or configuration process, the confluence of the system of rules also plays a major role: irrespective of the order of the rule applications on an initial state, the result is always the same.

The current research encompasses the analysis of existing rule-based systems, the development of a suitable formal language and the development of automatic methods for calculation of properties.

Ching Hoo Tang

RG. 1 Automation of Logic
Phone +49 681 9325-2919
Email chtang@mpi-inf.mpg.de

Christoph Weidenbach

RG. 1 Automation of Logic
Phone +49 681 9325-2900
Email weidenbach@mpi-inf.mpg.de