Towards Engineer-to-order Product Configuration

Constraint satisfaction problem (CSP) paradigm has proved highly successful for solving product configuration problems, particularly for build-to-order configuration in determining combination of different predefined component types. However, many real world customized products deal with engineer-to-order configuration, where product configuration is generated not only from combinations of different types of components, but also from varying different design parameters to satisfy pre-defined design constraints. The challenge posted by engineer-to-order configuration stems from n-ary constraints with continuous variables. Consequently, incorporating n-ary constraints and continuous variables into constraint satisfaction paradigm typically involves expensive search. Therefore, effective search strategies along with a modeling approach supporting the search strategies are the main issues in solving engineer-to-order configuration. In this paper, we present an approach to engineer-to-order product configuration where configuration problems are represented as CSP with n-ary constraints and variables with both discrete and continuous domains. We extended CSP by introducing two types of variables, where controllable variables can be independently assigned with values and dependable variables are derived based on those controllable variables. By limiting it into a set of controllable variables, search space is considerably reduced, and therefore, search efficiency is significant improved. The search algorithm is based on repair strategy in which min-conflicts heuristic is applied along with backtracking search. A case study indicated that this approach could effectively solve engineer-to-order product configuration by reducing its search space.