Abstract
Dynamic constraint satisfaction problem (DCSP) solving is one of the most important methods for solving various kinds of synthesis tasks, such as configuration. Todays configurators are standalone systems not supporting distributed configuration problem solving functionality. However, supply chain integration of configurable products requires the integration of configuration systems of different manufacturers, which jointly offer product solutions to their customers. As a consequence, we need problem solving methods that enable the computation of such configurations by several distributed configuration agents. Therefore, one possibility is the extension of the configuration problem from a dynamic constraint satisfaction representation to distributed dynamic constraint satisfaction (DDCSP). In this paper we will contribute to this challenge by formalizing the DDCSP and by presenting a complete and sound algorithm for solving distributed dynamic constraint satisfaction prob- lems. This algorithm is based on asynchronous backtracking and enables strategies for exploiting conflicting requirements and design assumptions (i.e. learning additional constraints during search). The exploitation of these additional constraints is of particular interest for configuration be- cause the generation and the exchange of conflicting design assumptions based on nogoods can be easily integrated in existing configuration systems.