A genetic algorithm to generate and evaluate assembly plans

This paper describes a genetic algorithm that deals with the assembly planning problem. While most assembly planning systems use a cut-set method to generate assembly plans, we propose a new approach to this problem: we use a genetic algorithm that generates and evaluates assembly plans. This algorithm starts from a set of valid assembly plans proposed by an expert of the product. This set is the initial population of potential solutions. Each assembly plan is encoded into a chromosome, to be manipulated by genetic operators. A reproduction process uses these operators to produce new assembly plans from "parents" assembly plans. An evaluation function and a selection procedure retain the best plans that expand the population and serve for new generations. This algorithm can be used either to quickly generate a set of good assembly plans, or to search all the valid assembly plans of a product.