Applications of non-monotonic reasoning to automotive product configuration using answer set programming

Abstract

In automotive industry, validation and maintenance of product configuration data is a complex task. Both orders from the customers and new product line designs from the R&D department are subject to a set of configuration rules to be satisfied. In this work, non-monotonic computational logic, answer set programming in particular, is applied to industrial-scale automotive product configuration problems. This methodology provides basic validation of the product configuration documentation and validation of single product orders, where Reiter style diagnosis provides minimal changes needed to correct an invalid order or a product configuration rule set. In addition, a method for discovering groups of product configuration variables that are strongly related can be obtained by small modification of the basic logic program, and by the usage of cautious and brave reasoning methods. As a result, options that are used in every, or respectively in no configuration, can easily be identified, as well as groups of options that are always used together or not at all. Finally it is possible to single out mandatory and obsolete options, relative to a preselected set of included or excluded options. Experimental results on an industrial dataset show applicability, example results, and computational feasibility with computation times on the order of seconds using a state-of-the-art answer set solver on standard PC hardware.