.png){kind=icon}
.png){kind=icon}
Return
Tri-objective distributed assembly flow shop scheduling with batch delivery: Indicator-Driven and reinforcement learning-enhanced artificial bee colony algorithms
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2026.114943.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In practical scenarios, products must be delivered promptly to customer when processing and assembly are completed. However, the existing literature on this topic is comparatively insufficient. This work investigates distributed assembly flow shop scheduling problems (DAFSP) with batch delivery. First, a mathematical model is firstly developed with minimizing three objectives, total energy consumption, total tardiness cost, and the sum of the maximum assembly completion time and the total delivery time, simultaneously. Second, the artificial bee colony (ABC) algorithm is improved to solve the concerned problems. An improved heuristic is designed to initialize population. Based on the characteristics of the coupled problems, two variable neighborhood search operators and three speed adjustment strategies are designed to enhance the algorithms’ performance. Third, two reinforcement learning algorithms are developed to choose the appropriate search operators or speed adjustment strategies during iterations. Two indicators, diversity and convergence, are employed to design the state-action pairs for precisely guiding the search direction. In addition, a reward scheme is developed to balance convergence and diversity, which linearly correlates with performance improvement. Further, two indicators, crowding distance and hypervolume contribution, are employed to design a diversity maintenance strategy, which evaluates the distribution of solutions and guides the search toward more promising regions. Finally, experimental analysis verifies the effectiveness of the improvement strategies. The results and discussions show that the improved ABC with Q-learning based strategies has the best performance among seven compared algorithms.
Keywords:
distributed assembly flow shop scheduling
batch delivery
tri-objective optimization
artificial bee colony algorithm
reinforcement learning
Journal
IF:
8
Papers:
5.3K
Citations:
3.5W
Organization
View more organizations