Smart outbound scheduling for textile warehousing: an integrated ALNS optimisation algorithm and LLM-based fuzzy decision-making framework

Order heterogeneity, multi-stage operations, and multi-production line coordination introduce considerable complexity to multi-Stock-Keeping Unit (SKU) outbound scheduling in smart textile warehouses. To address these challenges, we propose an integrated intelligent scheduling framework that combines advanced optimisation with human-centric decision support. The core of the framework is a Mixed-Integer Linear Programming (MILP) model that captures key practical constraints, such as workload balancing across production lines and penalties for order tardiness. To efficiently solve this model, we develop a tailored Adaptive Large Neighborhood Search (ALNS) algorithm featuring customised destroy-repair operators, an adaptive operator selection mechanism, and a simulated annealing-based acceptance criterion to enhance solution quality and convergence speed. Beyond algorithmic optimisation, the framework incorporates an interactive decision-making module that integrates T-Spherical Fuzzy Sets (T-SFS) with Large Language Models (LLMs). This module facilitates human-AI collaboration by capturing and processing decision-makers’ subjective preferences, allowing for flexible, preference-aware scheduling outcomes. Extensive computational experiments on benchmark instances demonstrate the superiority of our approach over traditional heuristics and metaheuristics in both performance and adaptability. The proposed framework offers a comprehensive solution for intelligent, responsive outbound scheduling in the textile warehousing domain.