Comparison of MILP and CP models for balancing partially automated assembly lines

Abstract

Abstract The objective of Assembly Line Balancing (ALB) is to find the proper assignment of tasks to workstations, taking into consideration various types of constraints and defined management goals. Early research in the field focused on solving the Simple Assembly Line Balancing problem, a basic simplified version of the general problem. As the production environment became more complex, several new ALB problem types appeared, and almost all ALB problems are NP-hard, meaning that finding a solution requires a lot of time, resources, and computational power. Methods with custom-made algorithms and generic approaches have been developed for solving these problems. While custom-made algorithms are generally more efficient, generic approaches can be more easily extended to cover other variations of the problem. Over the past few decades, automation has played an increasingly important role in various operations, although complete automation is often not possible. As a result, there is a growing need for partially automated assembly line balancing models. In these circumstances, the flexibility of a generic approach is essential. This paper compares two generic approaches: mixed integer linear programming (MILP) and constraint programming (CP), for two types of partially automated assembly line balancing problems. While CP is relatively slower in solving the simpler allocation problems, it is more efficient than MILP when an increased number of constraints is applied to the ALB and an allocation and scheduling problem needs to be solved.