A method to balancing robotic mixed-model assembly lines: Practical constraints, computational challenges, and performance estimation

This paper considers a robotic assembly line with various practical constraints. These have been previously modeled mathematically, but extending it to consider mixed-model production presents two challenges: computational costs increase substantially with problem size, and devising a theoretical model for practical performance is difficult within mathematical programming. This paper seeks to bridge those gaps,by proposing a method based on fixing, constraining, and optimizing mixed-integer programming instances. Simulations and linear relaxations are used to measure performance and estimate room for improvement. The resulting solution for the large-size industrial case study reached approximately 5% better throughput than a mixed- model benchmark approach, which represents around 85% of the estimated improvement potential for starting from that initial solution. Furthermore, an exhaustive set of tests was performed to demonstrate the proposed method's efficacy. Hence, the method managed to optimize a rather challenging computational problem that combines the complexity of relevant practical constraints with theoretical difficulties in estimating the performance of solutions.