Track: Graduate Student Paper Competition

Abstract

Who has never wanted to own a particular product, such as shoes, smart-phone, cosmetic, car, etc., specially designed for him/her, perfectly suited to his/her desires, and affordable ? For several decades now, customers want to bring a personal touch to their products to make them special and unique. To meet this demand of personalization, companies nowadays no longer offer standard products, but more and more personalizable ones. Thanks to the Web technologies, this personalization is done directly and interactively online. Customers can play with the range of choices and options offered by companies: they can assemble, cut, color, choose, ... visualize the result of their desires and ultimately order it, in a few minutes with a few clicks.  This concept of personalization or configuration of products consists in assembling modules or predefined components, to produce a unique and specific product. For businesses, this is a way to offer personalized products to stand out from the competition and build customers’ loyalty through more accurately reflecting their tastes and needs. The major difficulty of personalization or configuration of products lies in managing the diversity offered to customers: how to be sure that all the combinations of choices, variants and options offered to the customers are achievable in a reasonable time and affordable price ? Constraint satisfaction problems or CSP are very often used   to formalize product configuration problems in both research and industry. CSP formalize relevant knowledge through variables, each one associated to a definition domain, linked by constraints, limiting the combinations of their permissible values. Thus, CSP makes it possible  to describe exhaustively the solution space, corresponding to a set of all possible products. Two different  methods of processing CSP allow to exploit the generic models in an interactive way: problem filtering methods (reasoning directly on the CSP network and removing inconsistent values) and solution   filtering methods (reasoning on a representation of the solution space in the form of a compiled graph). Both of the methods have advantages and drawbacks in online product configuration. This paper aims at putting the first ideas on the joint use of these two methods in the same interactive configuration problem.