Exploring Efficiency and Solution Diversity for Type-E Assembly Line Balancing Problem

Abstract

This paper explores AI-driven optimization strategies for solving the Type-E Simple Assembly Line Balancing Problem (SALB-E), a complex combinatorial problem where both the number of stations (K) and cycle time (C) must be optimized to maximize line efficiency (E). We evaluate three approaches: Random Grid Search (RGS), a lightweight randomized baseline; Stochastic Hill Climbing (SHC), with tailored neighborhood operators; and a novel application of Quality Diversity Search (QDS), which employs the MAP-Elites framework to explore the trade-off space between K and C. QDS exemplifies a new class of algorithms that combine optimization with solution space illumination, producing diverse, high-performing configurations. Experiments on nine benchmark datasets show that QDS consistently outperforms traditional methods in robustness and insight generation, while SHC and RGS remain effective for simpler instances. These results highlight the promise of diversityaware approaches in AI-based optimization for complex assembly system design.

Interactive demo

A looping preview of the line-balancing model: an artificial immune system assigns precedence-constrained tasks to workstations under a cycle-time limit, rebalancing to raise line efficiency.